<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Great Work</title>
    <description>The latest articles on DEV Community by Great Work (@greatworkco).</description>
    <link>https://dev.to/greatworkco</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4216644%2F8335af88-5437-4ccc-9be1-47488948f8a7.png</url>
      <title>DEV Community: Great Work</title>
      <link>https://dev.to/greatworkco</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/greatworkco"/>
    <language>en</language>
    <item>
      <title>How to fix "No module named 'mcp.server.fastmcp'"</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Wed, 18 Nov 2026 22:47:07 +0000</pubDate>
      <link>https://dev.to/greatworkco/how-to-fix-no-module-named-mcpserverfastmcp-2p4b</link>
      <guid>https://dev.to/greatworkco/how-to-fix-no-module-named-mcpserverfastmcp-2p4b</guid>
      <description>&lt;p&gt;&lt;em&gt;Originally published at &lt;a href="https://runheed.com/errors/no-module-named-mcp-server-fastmcp" rel="noopener noreferrer"&gt;runheed.com&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Raised when Python imports your MCP server with mcp 2.0 or later installed (2.0 shipped on July 28, 2026; 2.3.0 is current). The class FastMCP is now MCPServer and every module under mcp.server.fastmcp moved to mcp.server.mcpserver. The SDK raises this ModuleNotFoundError with a hint that names the new import and the migration guide. A server run by Claude Desktop, Claude Code or Cursor dies at startup, so the client only reports that the connection closed.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Traceback (most recent call last):
  File "server.py", line 2, in &amp;lt;module&amp;gt;
    from mcp.server.fastmcp import FastMCP
ModuleNotFoundError: No module named 'mcp.server.fastmcp'. This is mcp 2.x, where FastMCP was renamed to MCPServer (from mcp.server.mcpserver import MCPServer) and other APIs changed; see the migration guide at https://py.sdk.modelcontextprotocol.io/v2/migration/#fastmcp-renamed-to-mcpserver or pin 'mcp&amp;lt;2' to keep running v1 code.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  The fix
&lt;/h2&gt;

&lt;p&gt;Change the import to from mcp.server.mcpserver import MCPServer and build the server with MCPServer("name"). If you are not ready to migrate, pin the dependency to mcp&amp;lt;2 so a fresh install doesn't pick up 2.x. Note that the separate fastmcp package (from fastmcp import FastMCP) is a different project and is not affected.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Run pip show mcp (or uv pip show mcp) to confirm the installed version is 2.x.&lt;/li&gt;
&lt;li&gt;Replace from mcp.server.fastmcp import FastMCP with from mcp.server.mcpserver import MCPServer, Context.&lt;/li&gt;
&lt;li&gt;Replace FastMCP("name") with MCPServer("name"), and ctx.fastmcp with ctx.mcp_server.&lt;/li&gt;
&lt;li&gt;Pass everything after the name by keyword: 2.x added title and description to the positional parameters, so a positional instructions string silently becomes the title.&lt;/li&gt;
&lt;li&gt;Move host and port from the constructor to run(). Or pin mcp&amp;lt;2 in requirements.txt or pyproject.toml and migrate later.
&lt;/li&gt;
&lt;/ol&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight python"&gt;&lt;code&gt;&lt;span class="c1"&gt;# mcp 1.x&lt;/span&gt;
&lt;span class="c1"&gt;# from mcp.server.fastmcp import FastMCP&lt;/span&gt;
&lt;span class="c1"&gt;# mcp = FastMCP("orders", "You look up orders.")&lt;/span&gt;

&lt;span class="c1"&gt;# mcp 2.x&lt;/span&gt;
&lt;span class="k"&gt;from&lt;/span&gt; &lt;span class="n"&gt;mcp&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;server&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;mcpserver&lt;/span&gt; &lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="nc"&gt;MCPServer&lt;/span&gt;

&lt;span class="n"&gt;mcp&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nc"&gt;MCPServer&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"orders"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;instructions&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;&lt;span class="s2"&gt;"You look up orders."&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;

&lt;span class="nd"&gt;@mcp&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;tool&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;span class="k"&gt;def&lt;/span&gt; &lt;span class="nf"&gt;get_order&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;order_id&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kt"&gt;str&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;-&amp;gt;&lt;/span&gt; &lt;span class="kt"&gt;str&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="s2"&gt;"""Look up an order."""&lt;/span&gt;
    &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nf"&gt;lookup&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;order_id&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;

&lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="n"&gt;__name__&lt;/span&gt; &lt;span class="o"&gt;==&lt;/span&gt; &lt;span class="s2"&gt;"__main__"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt;
    &lt;span class="n"&gt;mcp&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;run&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Why it happens
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;An unpinned dependency picked up mcp 2.x.&lt;/strong&gt; requirements.txt or pyproject.toml lists mcp with no upper bound, so a new environment, a CI build or uvx installs the latest major version.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Tutorial or template code written for 1.x.&lt;/strong&gt; Most published examples still import FastMCP from mcp.server.fastmcp.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Submodule imports.&lt;/strong&gt; Imports such as mcp.server.fastmcp.prompts or mcp.server.fastmcp.utilities fail the same way; they live under mcp.server.mcpserver now.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Also worth knowing
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;MCP_* environment variables and .env files are no longer read by MCPServer in 2.x: pass those settings explicitly.&lt;/li&gt;
&lt;li&gt;A server that crashes on import is reported by the client as MCP error -32000: Connection closed (from a Python client: MCPError: Connection closed inside an ExceptionGroup).&lt;/li&gt;
&lt;li&gt;After the upgrade, a tool that raises reaches the client only as "Error executing tool "; the cause is logged on the server.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Same error, other wording
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;code&gt;ModuleNotFoundError: No module named 'mcp.server.fastmcp.prompts'&lt;/code&gt;&lt;/li&gt;
&lt;li&gt;&lt;code&gt;ImportError: cannot import name 'FastMCP' from 'mcp.server'&lt;/code&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Sources
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://py.sdk.modelcontextprotocol.io/v2/migration/#fastmcp-renamed-to-mcpserver" rel="noopener noreferrer"&gt;MCP Python SDK v2 migration guide: FastMCP renamed to MCPServer&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://pypi.org/project/mcp/#history" rel="noopener noreferrer"&gt;mcp on PyPI (release history)&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>python</category>
      <category>mcp</category>
      <category>ai</category>
      <category>debugging</category>
    </item>
    <item>
      <title>See when Claude Code is done or needs you, in your Mac menu bar</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Wed, 18 Nov 2026 16:17:05 +0000</pubDate>
      <link>https://dev.to/greatworkco/see-when-claude-code-is-done-or-needs-you-in-your-mac-menu-bar-4oko</link>
      <guid>https://dev.to/greatworkco/see-when-claude-code-is-done-or-needs-you-in-your-mac-menu-bar-4oko</guid>
      <description>&lt;p&gt;&lt;em&gt;This guide first appeared on &lt;a href="https://multimeterapp.com/guides/claude-code-notifications" rel="noopener noreferrer"&gt;multimeterapp.com&lt;/a&gt;. Multimeter is a Mac menu bar app that turns any number into a live meter.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Start a long task, switch to something else, and stop checking the terminal. Claude Code runs hooks at fixed moments: when you send a prompt, after each tool, when it asks for permission and when it finishes. Four one-line hooks send a number to Multimeter at those moments, and the Traffic Light face shows it: yellow while Claude is working, green when it’s done, red when it’s waiting for you.&lt;/p&gt;

&lt;h2&gt;
  
  
  What you need
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;a href="https://multimeterapp.com/get-the-app" rel="noopener noreferrer"&gt;Multimeter&lt;/a&gt; for macOS 13 or later, $19.99 once.&lt;/li&gt;
&lt;li&gt;Claude Code, any plan or sign-in.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Set it up
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Add these hooks to &lt;code&gt;~/.claude/settings.json&lt;/code&gt; (merge the &lt;code&gt;hooks&lt;/code&gt; part with what’s already there). Claude Code picks them up for new sessions:
&lt;/li&gt;
&lt;/ol&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight json"&gt;&lt;code&gt;   &lt;span class="p"&gt;{&lt;/span&gt;
     &lt;span class="nl"&gt;"hooks"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
       &lt;span class="nl"&gt;"UserPromptSubmit"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;
         &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"hooks"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"type"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"command"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"command"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"open -g 'multimeter://set?key=claude&amp;amp;value=1&amp;amp;min=0&amp;amp;max=2&amp;amp;label=CLAUDE&amp;amp;formatted=working'"&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
       &lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
       &lt;span class="nl"&gt;"PostToolUse"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;
         &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"hooks"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"type"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"command"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"command"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"open -g 'multimeter://set?key=claude&amp;amp;value=1&amp;amp;min=0&amp;amp;max=2&amp;amp;label=CLAUDE&amp;amp;formatted=working'"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"async"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
       &lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
       &lt;span class="nl"&gt;"Notification"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;
         &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"matcher"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"permission_prompt"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"hooks"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"type"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"command"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"command"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"open -g 'multimeter://set?key=claude&amp;amp;value=2&amp;amp;min=0&amp;amp;max=2&amp;amp;label=CLAUDE&amp;amp;formatted=needs%20you'"&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
       &lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
       &lt;span class="nl"&gt;"Stop"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;
         &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"hooks"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nl"&gt;"type"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"command"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nl"&gt;"command"&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"open -g 'multimeter://set?key=claude&amp;amp;value=0&amp;amp;min=0&amp;amp;max=2&amp;amp;label=CLAUDE&amp;amp;formatted=done'"&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
       &lt;span class="p"&gt;]&lt;/span&gt;
     &lt;span class="p"&gt;}&lt;/span&gt;
   &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;ol&gt;
&lt;li&gt;In Multimeter, add a meter with the &lt;strong&gt;Sent Value&lt;/strong&gt; source and the key &lt;code&gt;claude&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Pick the &lt;strong&gt;Traffic Light&lt;/strong&gt; face. Set alerts to High values: warning at 1, critical at 2, and turn on &lt;strong&gt;Notify me when it goes critical&lt;/strong&gt; to get a Mac notification when Claude asks for permission. The hooks send the range (0 to 2), the label and a word for the meter’s menu.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Multimeter settings
&lt;/h2&gt;

&lt;p&gt;In &lt;strong&gt;Customize This Meter…&lt;/strong&gt; (from the meter’s menu):&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Settings for this meter&lt;/em&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Setting&lt;/th&gt;
&lt;th&gt;Value&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Source&lt;/td&gt;
&lt;td&gt;Sent Value&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Key&lt;/td&gt;
&lt;td&gt;&lt;code&gt;claude&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Range&lt;/td&gt;
&lt;td&gt;0 to 2, sent by the hooks&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Alerts&lt;/td&gt;
&lt;td&gt;High values: warning 1, critical 2, notify on critical&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Label&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;CLAUDE&lt;/code&gt;, sent by the hooks&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Face&lt;/td&gt;
&lt;td&gt;Traffic Light, Mood or Label &amp;amp; Value&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;strong&gt;Tested:&lt;/strong&gt; Run on November 18, 2026 with Claude Code in headless mode (&lt;code&gt;claude -p&lt;/code&gt;) and these hooks writing their numbers to a file: sending a prompt gave 1, running a tool gave 1, finishing gave 0. A permission prompt can’t happen in headless mode, so the red step wasn’t part of that run.&lt;/p&gt;

&lt;p&gt;More on this source, including every format it reads, in the &lt;a href="https://multimeterapp.com/guides/url-scheme" rel="noopener noreferrer"&gt;Sent Value guide&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Limits
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Multimeter sends at most one notification per meter every 15 minutes, and only when the meter turns red.&lt;/li&gt;
&lt;li&gt;One meter follows every Claude Code session on your Mac: the last session to change state sets the light.&lt;/li&gt;
&lt;li&gt;The hooks open a multimeter:// link in the background (&lt;code&gt;open -g&lt;/code&gt;), which takes a few milliseconds and doesn’t bring Multimeter forward.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Variations
&lt;/h2&gt;

&lt;h3&gt;
  
  
  A notification when it finishes instead
&lt;/h3&gt;

&lt;p&gt;Swap the numbers: send &lt;code&gt;value=2&lt;/code&gt; from the &lt;strong&gt;Stop&lt;/strong&gt; hook and &lt;code&gt;value=0&lt;/code&gt; from &lt;strong&gt;Notification&lt;/strong&gt;. The light then turns red when Claude is done, and the critical alert tells you.&lt;/p&gt;

&lt;h3&gt;
  
  
  One meter per project
&lt;/h3&gt;

&lt;p&gt;Put the hooks in a project’s &lt;code&gt;.claude/settings.json&lt;/code&gt; with its own key, like &lt;code&gt;key=claude-api&lt;/code&gt;, and give each project its own meter.&lt;/p&gt;

&lt;h3&gt;
  
  
  Usage too
&lt;/h3&gt;

&lt;p&gt;Pair it with the &lt;a href="https://multimeterapp.com/guides/claude-code-usage" rel="noopener noreferrer"&gt;Claude Code usage&lt;/a&gt; recipe to keep your 5-hour limit next to the light.&lt;/p&gt;

&lt;h2&gt;
  
  
  Troubleshooting
&lt;/h2&gt;

&lt;p&gt;&lt;em&gt;What the meter says, why, and what to do&lt;/em&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Symptom&lt;/th&gt;
&lt;th&gt;Cause&lt;/th&gt;
&lt;th&gt;Fix&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;The light never changes&lt;/td&gt;
&lt;td&gt;The hooks aren’t loaded: they apply to sessions started after you saved settings.json, or the JSON has a typo.&lt;/td&gt;
&lt;td&gt;Start a new session and run &lt;code&gt;/hooks&lt;/code&gt; in Claude Code to see them. Check the file with &lt;code&gt;jq . ~/.claude/settings.json&lt;/code&gt;.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;The meter says to give it a key&lt;/td&gt;
&lt;td&gt;The meter’s key doesn’t match the hooks.&lt;/td&gt;
&lt;td&gt;In Customize, set the key to &lt;code&gt;claude&lt;/code&gt; exactly.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;It stays red after you approved&lt;/td&gt;
&lt;td&gt;The next hook hasn’t run yet.&lt;/td&gt;
&lt;td&gt;It turns yellow after the next tool finishes, and green when Claude stops.&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  Questions
&lt;/h2&gt;

&lt;p&gt;No. Sent Value meters work in every version, and nothing here uses the Script source.&lt;/p&gt;

&lt;p&gt;No. Each hook opens one multimeter:// link in the background, and the hook after each tool runs asynchronously, so Claude doesn’t wait for it.&lt;/p&gt;

&lt;p&gt;Multimeter posts a regular Mac notification when the meter goes critical; its sound follows your notification settings for Multimeter in System Settings.&lt;/p&gt;

&lt;h3&gt;
  
  
  Related
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/claude-code-usage" rel="noopener noreferrer"&gt;Show your Claude Code usage in your Mac menu bar&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/github-actions-status" rel="noopener noreferrer"&gt;Show your GitHub Actions status in your Mac menu bar&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/url-scheme" rel="noopener noreferrer"&gt;Send a number to your menu bar with a link&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/faces/traffic-light" rel="noopener noreferrer"&gt;The Traffic Light face&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/developers" rel="noopener noreferrer"&gt;All developers recipes&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides" rel="noopener noreferrer"&gt;Every guide&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>claude</category>
      <category>macos</category>
      <category>productivity</category>
      <category>tutorial</category>
    </item>
    <item>
      <title>What does error tracking cost at 50k, 250k and 1M events a month?</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Mon, 16 Nov 2026 22:47:02 +0000</pubDate>
      <link>https://dev.to/greatworkco/what-does-error-tracking-cost-at-50k-250k-and-1m-events-a-month-3d10</link>
      <guid>https://dev.to/greatworkco/what-does-error-tracking-cost-at-50k-250k-and-1m-events-a-month-3d10</guid>
      <description>&lt;p&gt;&lt;em&gt;Originally published at &lt;a href="https://runheed.com/guides/error-tracking-pricing-compared" rel="noopener noreferrer"&gt;runheed.com&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;At 50,000 errors a month the entry plans cost $20 to $29. At about 250,000 they cost $65 to $79, and at 1 million $176 to $225, depending on vendor and billing. Price alone separates them less than retention, what happens when you go over, and whether your coding agent can read and resolve the errors. Runheed is flat: $29, $79 and $199 a month, and events over the limit are rejected, never billed.&lt;/p&gt;

&lt;h2&gt;
  
  
  The prices at three volumes
&lt;/h2&gt;

&lt;p&gt;Each figure comes from the vendor's own pricing page on November 11, 2026, for the cheapest paid plan that covers the volume. Where a vendor sells in steps, the price is for the first step at or above the volume, shown in brackets.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Tool and plan&lt;/th&gt;
&lt;th&gt;50k errors/month&lt;/th&gt;
&lt;th&gt;About 250k&lt;/th&gt;
&lt;th&gt;1M&lt;/th&gt;
&lt;th&gt;Retention&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Runheed (flat, monthly)&lt;/td&gt;
&lt;td&gt;$29 (Solo)&lt;/td&gt;
&lt;td&gt;$79 (Pro, 250k)&lt;/td&gt;
&lt;td&gt;$199 (Scale)&lt;/td&gt;
&lt;td&gt;30 / 90 / 180 days&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Sentry Team (billed annually, reserved volume)&lt;/td&gt;
&lt;td&gt;$26&lt;/td&gt;
&lt;td&gt;about $67&lt;/td&gt;
&lt;td&gt;about $186&lt;/td&gt;
&lt;td&gt;Up to 90 days&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Sentry Team (billed annually, pay-as-you-go over 50k)&lt;/td&gt;
&lt;td&gt;$26&lt;/td&gt;
&lt;td&gt;about $77&lt;/td&gt;
&lt;td&gt;about $225&lt;/td&gt;
&lt;td&gt;Up to 90 days&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Honeybadger Team (monthly)&lt;/td&gt;
&lt;td&gt;$26&lt;/td&gt;
&lt;td&gt;$74 (300K)&lt;/td&gt;
&lt;td&gt;$176&lt;/td&gt;
&lt;td&gt;90 days&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Rollbar Essentials (monthly)&lt;/td&gt;
&lt;td&gt;$29&lt;/td&gt;
&lt;td&gt;$79 (300k)&lt;/td&gt;
&lt;td&gt;$199&lt;/td&gt;
&lt;td&gt;90 days&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;BugSnag Select (list price)&lt;/td&gt;
&lt;td&gt;$20&lt;/td&gt;
&lt;td&gt;$65 (300k)&lt;/td&gt;
&lt;td&gt;$200&lt;/td&gt;
&lt;td&gt;60 days&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Sentry's figures add its published per-error rates for the Team plan to the $26 base, which includes 50k errors: reserved volume is prepaid at a discount, pay-as-you-go is billed after the fact. Sentry rounds those rates, so treat the totals as close, not exact. Honeybadger's figures are its Team base plus the error add-on from its pricing calculator. Rollbar and BugSnag publish a price for each step.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the price does not show
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Going over: Sentry bills extra volume from a pay-as-you-go budget you set; the stepped plans move you up a step. Runheed rejects events past the plan limit and tells you, so the bill never changes.&lt;/li&gt;
&lt;li&gt;AI fixing: Sentry sells its AI debugging agent, Seer, on top of the plan ($40 per active contributor a month until December 8, 2026, then usage-based credits). With Runheed your own coding agent does the fix over MCP, with no add-on.&lt;/li&gt;
&lt;li&gt;Agent access: all five offer an MCP server, so the question is what the agent can do with it. The comparison of error tracking MCP servers covers that tool by tool.&lt;/li&gt;
&lt;li&gt;Everything else: Sentry, Honeybadger and BugSnag bundle tracing, uptime, logs or mobile stability. Runheed does errors only, with Node.js and Python SDKs and no browser or mobile SDK yet.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  How to pick on cost
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Count your real error volume for a month (most trackers show it on the usage page). Expect spikes after a bad deploy.&lt;/li&gt;
&lt;li&gt;Price the step above your volume, not the one at it: a spike to the next step is where the bills differ.&lt;/li&gt;
&lt;li&gt;Decide what you want to happen at the limit: pay for the extra, or drop it.&lt;/li&gt;
&lt;li&gt;Add what you would pay for AI fixing, if you want it, on top of the plan.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Which error tracker is cheapest?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;At these three volumes, BugSnag Select is lowest at 50k and about 250k, and Honeybadger Team is lowest at 1M. The gaps are small, a few dollars to a few tens of dollars a month, so retention and overage rules usually matter more.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Does Runheed charge for overage?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;No. Each plan has a monthly event limit (50,000, 250,000 or 1,000,000). Events past it are rejected, not billed, and you can move to a bigger plan any time.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Are these prices current?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;They were read from each vendor's pricing page on November 11, 2026. Vendors change prices; check their pages before you buy. Sentry's Seer pricing changes on December 8, 2026.&lt;/p&gt;

&lt;h2&gt;
  
  
  Related
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://runheed.com/compare" rel="noopener noreferrer"&gt;Runheed compared with Sentry, Honeybadger, Rollbar and BugSnag&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://runheed.com/compare/mcp-servers" rel="noopener noreferrer"&gt;Error tracking MCP servers compared&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://runheed.com/compare/sentry" rel="noopener noreferrer"&gt;Runheed vs Sentry&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://runheed.com/pricing" rel="noopener noreferrer"&gt;Runheed pricing (7-day trial)&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>webdev</category>
      <category>devops</category>
      <category>monitoring</category>
      <category>programming</category>
    </item>
    <item>
      <title>Show your unread Mail count in your Mac menu bar</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Sun, 15 Nov 2026 16:17:06 +0000</pubDate>
      <link>https://dev.to/greatworkco/show-your-unread-mail-count-in-your-mac-menu-bar-348</link>
      <guid>https://dev.to/greatworkco/show-your-unread-mail-count-in-your-mac-menu-bar-348</guid>
      <description>&lt;p&gt;&lt;em&gt;This guide first appeared on &lt;a href="https://multimeterapp.com/guides/mail-unread-count" rel="noopener noreferrer"&gt;multimeterapp.com&lt;/a&gt;. Multimeter is a Mac menu bar app that turns any number into a live meter.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Mail’s Dock badge is easy to miss when the Dock is hidden, and Mail’s window is usually behind everything else. A Script meter asks Mail with AppleScript how many messages in the Inbox are unread and puts the number next to the clock.&lt;/p&gt;

&lt;h2&gt;
  
  
  What you need
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;a href="https://multimeterapp.com/get-the-app" rel="noopener noreferrer"&gt;Multimeter&lt;/a&gt; for macOS 13 or later, $19.99 once.&lt;/li&gt;
&lt;li&gt;The direct download version of Multimeter from this site (the Script source isn’t in the Mac App Store version).&lt;/li&gt;
&lt;li&gt;Multimeter 1.1.1 or later. Earlier versions can’t ask macOS for permission to talk to other apps. Choose &lt;strong&gt;Check for Updates…&lt;/strong&gt; from any meter’s menu.&lt;/li&gt;
&lt;li&gt;Apple Mail with at least one account set up.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Set it up
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Here’s the command. It asks Mail for the unread count of the Inbox, the one at the top of Mail’s sidebar that combines your accounts, and prints a placeholder when Mail isn’t running:
&lt;/li&gt;
&lt;/ol&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;   osascript -e &lt;span class="s1"&gt;'if application "Mail" is running then'&lt;/span&gt; -e &lt;span class="s1"&gt;'tell application "Mail" to return unread count of inbox'&lt;/span&gt; -e &lt;span class="s1"&gt;'else'&lt;/span&gt; -e &lt;span class="s1"&gt;'return "{\"value\": 0, \"formatted\": \"closed\"}"'&lt;/span&gt; -e &lt;span class="s1"&gt;'end if'&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;ol&gt;
&lt;li&gt;In Multimeter, choose &lt;strong&gt;Add Meter…&lt;/strong&gt; from any meter’s menu, pick &lt;strong&gt;Script&lt;/strong&gt; and paste the command.&lt;/li&gt;
&lt;li&gt;Set &lt;strong&gt;Refresh every&lt;/strong&gt; to 1 minute. Mail answers in well under a second, but there’s no reason to ask it more often, and 5 minutes is fine for a quieter inbox.&lt;/li&gt;
&lt;li&gt;Set the label to &lt;code&gt;MAIL&lt;/code&gt;, the range to 0 to 50, and alerts to High values with a warning at 10 and critical at 25 if you want the number to change color as mail piles up.&lt;/li&gt;
&lt;li&gt;With Mail open, wait for the first refresh. macOS asks once: &lt;strong&gt;“Multimeter” wants access to control “Mail”. Allowing control will provide access to documents and data in “Mail”, and to perform actions within that app.&lt;/strong&gt; Below it is Multimeter’s reason: “Multimeter runs your script meters, and scripts that read other apps (like Things, Reminders, Mail or Music) need your permission to control them.” Click &lt;strong&gt;Allow&lt;/strong&gt;. Until you do, the meter shows a dash; the number appears at the next refresh, or right away with &lt;strong&gt;Refresh Now&lt;/strong&gt; in the meter’s menu.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Multimeter settings
&lt;/h2&gt;

&lt;p&gt;In &lt;strong&gt;Customize This Meter…&lt;/strong&gt; (from the meter’s menu):&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Settings for this meter&lt;/em&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Setting&lt;/th&gt;
&lt;th&gt;Value&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Source&lt;/td&gt;
&lt;td&gt;Script&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Command&lt;/td&gt;
&lt;td&gt;the &lt;code&gt;osascript&lt;/code&gt; line above&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Refresh every&lt;/td&gt;
&lt;td&gt;1 minute or slower&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Range&lt;/td&gt;
&lt;td&gt;0 to 50&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Alerts&lt;/td&gt;
&lt;td&gt;High values: warning 10, critical 25 (optional)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Label&lt;/td&gt;
&lt;td&gt;&lt;code&gt;MAIL&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Face&lt;/td&gt;
&lt;td&gt;Label &amp;amp; Value or LCD Readout&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Width&lt;/td&gt;
&lt;td&gt;120 pt for Label &amp;amp; Value&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;strong&gt;Tested:&lt;/strong&gt; Set up on October 31, 2026 in Multimeter 1.1.1 on macOS 27, with one test mail account holding 7 unread messages. macOS showed the Automation prompt with Multimeter’s reason at the first refresh; after Allow the menu bar showed &lt;code&gt;MAIL 7&lt;/code&gt; in Label &amp;amp; Value and LCD Readout. After 2 messages were marked read it showed &lt;code&gt;MAIL 5&lt;/code&gt; at the next refresh. With Mail quit, both meters showed &lt;code&gt;closed&lt;/code&gt; and Mail stayed closed for the next two refreshes; reopening Mail brought back &lt;code&gt;MAIL 5&lt;/code&gt;. The command took about half a second when run in Terminal.&lt;/p&gt;

&lt;p&gt;Script meters need the direct download version of Multimeter from this site, which is the one on sale now. How commands run, the PATH they see and the 10 second limit are in the &lt;a href="https://multimeterapp.com/guides/script-meters" rel="noopener noreferrer"&gt;Script meters guide&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Limits
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Mail must be open. That’s on purpose: asking a closed Mail would launch it.&lt;/li&gt;
&lt;li&gt;The count is the Inbox only. Unread mail that rules moved to other mailboxes isn’t included; see the variation below for one mailbox.&lt;/li&gt;
&lt;li&gt;If Mail is set to show Categories, its Dock badge can count Primary only. This meter counts every unread message in the Inbox.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Variations
&lt;/h2&gt;

&lt;h3&gt;
  
  
  One account’s inbox
&lt;/h3&gt;

&lt;p&gt;Replace &lt;code&gt;inbox&lt;/code&gt; with the account’s mailbox. The account name is the one in Mail’s sidebar:&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Command&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight applescript"&gt;&lt;code&gt;&lt;span class="n"&gt;osascript&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="n"&gt;e&lt;/span&gt; &lt;span class="c1"&gt;'if application "Mail" is running then' -e 'tell application "Mail" to return unread count of mailbox "INBOX" of account "Work"' -e 'end if'&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  Notify me when mail arrives
&lt;/h3&gt;

&lt;p&gt;Set alerts to High values with a critical level of 1 and turn on &lt;strong&gt;Notify me when it goes critical&lt;/strong&gt;. Multimeter waits 15 minutes between notifications for the same meter, so a busy inbox won’t keep pinging you.&lt;/p&gt;

&lt;h2&gt;
  
  
  Troubleshooting
&lt;/h2&gt;

&lt;p&gt;&lt;em&gt;What the meter says, why, and what to do&lt;/em&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Symptom&lt;/th&gt;
&lt;th&gt;Cause&lt;/th&gt;
&lt;th&gt;Fix&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;The meter says closed&lt;/td&gt;
&lt;td&gt;Mail isn’t running. The command skips Mail when it’s closed so it never launches it.&lt;/td&gt;
&lt;td&gt;Open Mail. The number shows at the next refresh.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;The meter shows a dash while Mail is open&lt;/td&gt;
&lt;td&gt;Multimeter isn’t allowed to control Mail: you clicked Don’t Allow, or the question is still waiting on screen (the script stops after 10 seconds while it waits).&lt;/td&gt;
&lt;td&gt;Answer the prompt if it’s open. If you turned it down, open System Settings, &lt;strong&gt;Privacy &amp;amp; Security&lt;/strong&gt;, &lt;strong&gt;Automation&lt;/strong&gt;, click the arrow next to Multimeter and turn on Mail. The next refresh shows the number.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Works in Terminal, nothing in the meter&lt;/td&gt;
&lt;td&gt;Terminal and Multimeter each need their own permission. Allowing Terminal to control Mail doesn’t allow Multimeter.&lt;/td&gt;
&lt;td&gt;Look for Multimeter, not Terminal, under System Settings, Privacy &amp;amp; Security, Automation, and turn on Mail there.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;The script didn’t print a number&lt;/td&gt;
&lt;td&gt;The command failed or printed an error message instead.&lt;/td&gt;
&lt;td&gt;Run the exact command in Terminal. If it works there, check for tools that live outside the meter’s PATH and use their full path (&lt;code&gt;which tool&lt;/code&gt;).&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;The script took longer than 10 seconds&lt;/td&gt;
&lt;td&gt;A slow network call. Multimeter stops scripts after 10 seconds.&lt;/td&gt;
&lt;td&gt;Keep the &lt;code&gt;-m&lt;/code&gt; timeout on curl, and refresh less often.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Scripts need the version from multimeterapp.com&lt;/td&gt;
&lt;td&gt;This copy of Multimeter came from the Mac App Store, which can’t run scripts.&lt;/td&gt;
&lt;td&gt;Use the direct download version, or run the command from cron or launchd and point a &lt;a href="https://multimeterapp.com/guides/file-meters" rel="noopener noreferrer"&gt;File meter&lt;/a&gt; at its output.&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  Questions
&lt;/h2&gt;

&lt;p&gt;Script meters need the direct download version from multimeterapp.com, because the App Store sandbox doesn’t let apps run your commands. Web Address, File and Sent Value meters work in every version.&lt;/p&gt;

&lt;p&gt;No. The command asks Mail for one number, the unread count, and nothing else. macOS’s prompt describes the permission in general terms (“access to documents and data in Mail”) because it covers any script you might write; the one on this page reads only the count.&lt;/p&gt;

&lt;p&gt;Mail already knows the count for every account you’ve added, Gmail, iCloud, Exchange or your own server, so one line covers all of them without tokens or passwords.&lt;/p&gt;

&lt;h3&gt;
  
  
  Related
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/things-today-count" rel="noopener noreferrer"&gt;Show your Things Today count in your Mac menu bar&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/github-notifications" rel="noopener noreferrer"&gt;Show unread GitHub notifications in your Mac menu bar&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/script-meters" rel="noopener noreferrer"&gt;Show the output of a shell command in your menu bar&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/faces/label-value" rel="noopener noreferrer"&gt;The Label &amp;amp; Value face&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/productivity" rel="noopener noreferrer"&gt;All mail and tasks recipes&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides" rel="noopener noreferrer"&gt;Every guide&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>macos</category>
      <category>applescript</category>
      <category>productivity</category>
      <category>email</category>
    </item>
    <item>
      <title>Keep your Sentry SDK and send its errors to your coding agent</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Sat, 14 Nov 2026 22:47:05 +0000</pubDate>
      <link>https://dev.to/greatworkco/keep-your-sentry-sdk-and-send-its-errors-to-your-coding-agent-3d0c</link>
      <guid>https://dev.to/greatworkco/keep-your-sentry-sdk-and-send-its-errors-to-your-coding-agent-3d0c</guid>
      <description>&lt;p&gt;&lt;em&gt;Originally published at &lt;a href="https://runheed.com/guides/use-sentry-sdk-with-runheed" rel="noopener noreferrer"&gt;runheed.com&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Set your existing Sentry SDK's DSN to your Runheed project's Sentry SDK DSN, shaped https://@runheed.com/. No other code changes: exceptions, messages, stack frames, breadcrumbs, user, tags, release and environment arrive in Runheed, and your coding agent reads them over MCP.&lt;/p&gt;

&lt;h2&gt;
  
  
  Switch in three steps
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Create a project in Runheed (or have your agent call runheed_create_project). Copy its Sentry SDK DSN from the Projects page; runheed_list_projects shows it too.&lt;/li&gt;
&lt;li&gt;Replace the DSN your Sentry.init() reads, usually the SENTRY_DSN environment variable, with that value. Redeploy.&lt;/li&gt;
&lt;li&gt;Connect your coding agent to &lt;a href="https://runheed.com/api/mcp" rel="noopener noreferrer"&gt;https://runheed.com/api/mcp&lt;/a&gt; and ask it for the newest unresolved errors.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;code&gt;server.ts&lt;/code&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="o"&gt;*&lt;/span&gt; &lt;span class="k"&gt;as&lt;/span&gt; &lt;span class="nc"&gt;Sentry&lt;/span&gt; &lt;span class="k"&gt;from&lt;/span&gt; &lt;span class="s1"&gt;'@sentry/node'&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="c1"&gt;// Before: SENTRY_DSN=https://...@o123.ingest.sentry.io/456&lt;/span&gt;
&lt;span class="c1"&gt;// After:  SENTRY_DSN=https://&amp;lt;key&amp;gt;@runheed.com/&amp;lt;number&amp;gt;&lt;/span&gt;
&lt;span class="nc"&gt;Sentry&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;init&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;dsn&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nb"&gt;process&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;env&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="no"&gt;SENTRY_DSN&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;release&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nb"&gt;process&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;env&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="no"&gt;GIT_SHA&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Which SDKs work
&lt;/h2&gt;

&lt;p&gt;Every Sentry SDK sends the same envelope format, so any of them can report to Runheed. These were run end to end against Runheed on 2026-11-13:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Language&lt;/th&gt;
&lt;th&gt;SDK tested&lt;/th&gt;
&lt;th&gt;Setup page&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Node.js / TypeScript&lt;/td&gt;
&lt;td&gt;@sentry/node 11.12&lt;/td&gt;
&lt;td&gt;/docs#sentry-sdk&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Python&lt;/td&gt;
&lt;td&gt;sentry-sdk 2.75 (gzip, chained exceptions)&lt;/td&gt;
&lt;td&gt;/docs#sentry-sdk&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Go&lt;/td&gt;
&lt;td&gt;sentry-go 0.50&lt;/td&gt;
&lt;td&gt;/for/go&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Ruby and Rails&lt;/td&gt;
&lt;td&gt;sentry-ruby 7.2&lt;/td&gt;
&lt;td&gt;/for/ruby&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;PHP and Laravel&lt;/td&gt;
&lt;td&gt;sentry/sentry 4.37&lt;/td&gt;
&lt;td&gt;/for/php&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Java and Spring Boot&lt;/td&gt;
&lt;td&gt;sentry-java 8.16&lt;/td&gt;
&lt;td&gt;/for/java&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;.NET and ASP.NET Core&lt;/td&gt;
&lt;td&gt;Sentry 7.0&lt;/td&gt;
&lt;td&gt;/for/dotnet&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Browser (React, Vue, plain JS)&lt;/td&gt;
&lt;td&gt;@sentry/browser 11 in Chrome, cross-origin&lt;/td&gt;
&lt;td&gt;/for/react&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Framework integrations (sentry-rails, sentry-laravel, the Spring Boot starter, Sentry.AspNetCore) build on these SDKs and send the same events.&lt;/p&gt;

&lt;h2&gt;
  
  
  What comes across, and what does not
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;From the Sentry SDK&lt;/th&gt;
&lt;th&gt;In Runheed&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Exceptions and their chain (raise ... from, cause)&lt;/td&gt;
&lt;td&gt;Kept: the outermost exception is the error, the rest are listed as causes&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;captureMessage / capture_message&lt;/td&gt;
&lt;td&gt;Kept, as a Message error&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Stack frames with source context&lt;/td&gt;
&lt;td&gt;Kept, your own frames marked in-app, crash site first&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Breadcrumbs, user, request, tags, extra&lt;/td&gt;
&lt;td&gt;Kept&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;release and environment&lt;/td&gt;
&lt;td&gt;Kept, so the agent can ask what a release introduced&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;A custom fingerprint&lt;/td&gt;
&lt;td&gt;Used for grouping&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Minified browser frames&lt;/td&gt;
&lt;td&gt;Mapped to your source through the bundle's public source map&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Transactions, spans, profiles, replays, sessions, check-ins&lt;/td&gt;
&lt;td&gt;Accepted and dropped: Runheed tracks errors only&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Set tracesSampleRate (traces_sample_rate) to 0 so the SDK does not send performance data that would be dropped anyway.&lt;/p&gt;

&lt;h2&gt;
  
  
  Want Sentry and Runheed side by side?
&lt;/h2&gt;

&lt;p&gt;A Sentry SDK sends to one DSN. To keep Sentry for tracing and replay and also give your agent the errors, keep Sentry's DSN as it is and add the Runheed SDK next to it (Node.js and Python), or send a copy from your own code to the Runheed ingest endpoint. The /sentry-alternative page shows the side-by-side setup.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Do I need to change any code besides the DSN?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;No. Everything the SDK already captures keeps working. Optionally turn tracing off, since performance data is dropped.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Do past Sentry issues move over?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;No. Runheed starts grouping from the first event it receives.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;What happens at the monthly event limit?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Runheed answers 429 with Retry-After, which Sentry SDKs honor, so the SDK backs off instead of retrying in a loop. You are never billed for overage.&lt;/p&gt;

&lt;h2&gt;
  
  
  Related
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://runheed.com/docs#sentry-sdk" rel="noopener noreferrer"&gt;Docs: using a Sentry SDK&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://runheed.com/sentry-alternative" rel="noopener noreferrer"&gt;Sentry alternative for coding agents&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://runheed.com/guides/does-sentry-have-an-mcp-server" rel="noopener noreferrer"&gt;Does Sentry have an MCP server?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://runheed.com/for" rel="noopener noreferrer"&gt;Setup by framework&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>sentry</category>
      <category>ai</category>
      <category>devops</category>
      <category>tutorial</category>
    </item>
    <item>
      <title>Answer a blocked coding agent from your phone or any browser</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Sat, 14 Nov 2026 09:03:16 +0000</pubDate>
      <link>https://dev.to/greatworkco/answer-a-blocked-coding-agent-from-your-phone-or-any-browser-4l01</link>
      <guid>https://dev.to/greatworkco/answer-a-blocked-coding-agent-from-your-phone-or-any-browser-4l01</guid>
      <description>&lt;p&gt;&lt;em&gt;Originally published at &lt;a href="https://taskhold.com/guides/answer-a-blocked-coding-agent" rel="noopener noreferrer"&gt;taskhold.com&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Short answer: A terminal alert helps only if you're at the machine running the agent. To answer from anywhere, have the agent block the task with one specific question on a shared board, reply there from any browser, and let the agent read the reply when it next checks. Taskhold does this with taskhold_block and taskhold_updates, for any MCP agent.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;An agent that stops to ask something and waits in a terminal you aren't looking at costs you the whole wait, and an alert only helps if you're near that machine. This guide covers answering from somewhere else. The agent puts its question on Taskhold's live board, where a blocked task jumps to the top with a reply box, and picks up your answer the next time it checks. Try it in the &lt;a href="https://taskhold.com/demo/live" rel="noopener noreferrer"&gt;live board demo&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Getting the alert in the first place is a separate fix for each tool, with a guide for &lt;a href="https://taskhold.com/guides/claude-code-waiting-for-input" rel="noopener noreferrer"&gt;Claude Code&lt;/a&gt;, &lt;a href="https://taskhold.com/guides/codex-cli-waiting-for-approval" rel="noopener noreferrer"&gt;Codex CLI&lt;/a&gt;, &lt;a href="https://taskhold.com/guides/cursor-agent-waiting-for-input" rel="noopener noreferrer"&gt;Cursor&lt;/a&gt; and &lt;a href="https://taskhold.com/guides/gemini-cli-waiting-for-input" rel="noopener noreferrer"&gt;Gemini CLI or Antigravity CLI&lt;/a&gt;. The &lt;a href="https://taskhold.com/docs/slack-alerts" rel="noopener noreferrer"&gt;Slack alerts recipe&lt;/a&gt; and &lt;a href="https://taskhold.com/docs/claude-code-hooks" rel="noopener noreferrer"&gt;Claude Code hooks&lt;/a&gt; cover other ways to get told, and &lt;a href="https://taskhold.com/guides/monitor-coding-agent-progress" rel="noopener noreferrer"&gt;seeing what your agent is doing&lt;/a&gt; covers the rest of the board.&lt;/p&gt;

&lt;h2&gt;
  
  
  Answering from somewhere other than the terminal
&lt;/h2&gt;

&lt;p&gt;Terminal alerts point you at one machine. When you're away from it, or running agents in several tools, you need the question somewhere you can read and answer, and the agent needs a way to pick the answer up.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;The agent calls taskhold_block with a specific question, for example "Keep the v1 endpoints alive, or delete them in this PR?"&lt;/li&gt;
&lt;li&gt;The task moves to the top of Taskhold's live board under Needs you. With notifications on, the open board tab shows a desktop notification.&lt;/li&gt;
&lt;li&gt;You type the answer and press Send, or Send and unblock to send the task back to the queue. On a phone this is the same page.&lt;/li&gt;
&lt;li&gt;The agent calls taskhold_updates with its cursor between steps or at the start of the next session, reads your reply, and carries on, or takes another task while it waits.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Answering several questions at once
&lt;/h2&gt;

&lt;p&gt;When several agents are waiting, or one agent has a batch of decisions, your agent can link you to Taskhold's &lt;a href="https://taskhold.com/docs/links-for-agents#answer-decisions-on-one-page" rel="noopener noreferrer"&gt;decisions page&lt;/a&gt;. It lists only the open questions, each with the agent's choices, a None of these option and a notes field, and one Send answers everything you filled in. Anything you leave blank stays open. Copy answers gives you the same answers as text for an agent's chat.&lt;/p&gt;

&lt;h2&gt;
  
  
  Questions an agent can act on
&lt;/h2&gt;

&lt;p&gt;A good blocking question names the choice and the default. Put the rule in CLAUDE.md or AGENTS.md: "Can't continue? taskhold_block with a specific question, then check taskhold_updates for the answer."&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Vague: "Need clarification on the auth changes."&lt;/li&gt;
&lt;li&gt;Answerable: "Sessions expire after 24h today. Change to 7 days as the ticket says, or keep 24h for admin users?"&lt;/li&gt;
&lt;li&gt;Answerable: "STRIPE_TEST_KEY isn't set in .env. Add it, or should I mock Stripe in these tests?"&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;More patterns for wording questions and progress notes are in &lt;a href="https://taskhold.com/docs/agent-messages" rel="noopener noreferrer"&gt;agent messages&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Getting the alert first
&lt;/h2&gt;

&lt;p&gt;Claude Code, Codex, Cursor and Gemini CLI can each alert you on the machine running them, but whether an alert reaches you depends on your terminal, window focus and the kind of prompt. The settings and hooks that fix that are in &lt;a href="https://taskhold.com/guides/claude-code-waiting-for-input" rel="noopener noreferrer"&gt;Claude Code waits for input silently&lt;/a&gt;, &lt;a href="https://taskhold.com/guides/codex-cli-waiting-for-approval" rel="noopener noreferrer"&gt;Codex CLI waits for approval silently&lt;/a&gt;, &lt;a href="https://taskhold.com/guides/cursor-agent-waiting-for-input" rel="noopener noreferrer"&gt;Cursor agent waits for input silently&lt;/a&gt; and &lt;a href="https://taskhold.com/guides/gemini-cli-waiting-for-input" rel="noopener noreferrer"&gt;Gemini CLI waits for input silently&lt;/a&gt;. For a phone push whenever a task gets blocked, run the &lt;a href="https://taskhold.com/integrations/ntfy" rel="noopener noreferrer"&gt;ntfy script&lt;/a&gt; or the &lt;a href="https://taskhold.com/integrations/pushover" rel="noopener noreferrer"&gt;Pushover script&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Limits to know
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Taskhold doesn't wake a sleeping agent. The agent reads your reply the next time it calls taskhold_updates, which the snippet asks for between steps and at session start.&lt;/li&gt;
&lt;li&gt;Board notifications are browser notifications, shown while the live board is open in a tab. Taskhold doesn't send phone push notifications itself; the ntfy and Pushover scripts above do that from your own machine.&lt;/li&gt;
&lt;li&gt;Permission prompts inside Claude Code or Codex still need an answer in that tool. Taskhold carries questions about the work, not tool approvals.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Can I answer Claude Code's question from my phone without Taskhold?
&lt;/h3&gt;

&lt;p&gt;Yes, for Claude Code: start the session with Remote Control and answer in the Claude app, which can also send push notifications. Taskhold is for questions from any agent, collected on one board.&lt;/p&gt;

&lt;h3&gt;
  
  
  What happens to a blocked task if the session ends?
&lt;/h3&gt;

&lt;p&gt;Taskhold keeps it blocked with the question. Your reply waits on the task, and the next session reads it with taskhold_updates or taskhold_get before it picks up work.&lt;/p&gt;

&lt;h3&gt;
  
  
  Does the agent sit idle while it waits?
&lt;/h3&gt;

&lt;p&gt;That's up to your instructions. A common rule is to block the task, then take the next todo, so one question doesn't stop the whole queue.&lt;/p&gt;

&lt;h2&gt;
  
  
  Sources
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;a href="https://code.claude.com/docs/en/mobile" rel="noopener noreferrer"&gt;Claude Code docs: Claude Code on mobile&lt;/a&gt;, checked 2026-10-26&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Related
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/guides/openai-agents-sdk-progress" rel="noopener noreferrer"&gt;OpenAI Agents SDK progress updates for long runs&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/guides/monitor-coding-agent-progress" rel="noopener noreferrer"&gt;How to see what Claude Code is doing, live&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/guides/claude-code-waiting-for-input" rel="noopener noreferrer"&gt;Claude Code waits for input silently: the fix&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/guides/codex-cli-waiting-for-approval" rel="noopener noreferrer"&gt;Codex stuck on "Awaiting approval": the fix&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/guides/hand-tasks-to-your-coding-agent" rel="noopener noreferrer"&gt;How to hand tasks to your coding agent&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/guides/claude-md-task-workflow" rel="noopener noreferrer"&gt;A CLAUDE.md and AGENTS.md task workflow&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/for/codex" rel="noopener noreferrer"&gt;Codex CLI task list over MCP: watch tasks live&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/vs/claude-code-remote-control" rel="noopener noreferrer"&gt;Taskhold vs Claude Code Remote Control&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  About Taskhold
&lt;/h2&gt;

&lt;p&gt;Taskhold is a hosted task list shared by a person and their AI coding agents over MCP. Agents break tasks into steps and report progress, and the person watches a live board. $12/month or $120/year per person, 7-day trial with a card, no free plan. Made by Great Work LLC.&lt;/p&gt;

&lt;p&gt;To try it, add the MCP server &lt;a href="https://taskhold.com/api/mcp" rel="noopener noreferrer"&gt;https://taskhold.com/api/mcp&lt;/a&gt; to your agent. It connects with no sign-up, and the first time the agent uses Taskhold it gives you a link to start the trial.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Pricing: &lt;a href="https://taskhold.com/#pricing" rel="noopener noreferrer"&gt;https://taskhold.com/#pricing&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>ai</category>
      <category>productivity</category>
      <category>devtools</category>
      <category>tutorial</category>
    </item>
    <item>
      <title>Show your open browser tab count in your Mac menu bar</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Fri, 13 Nov 2026 16:17:05 +0000</pubDate>
      <link>https://dev.to/greatworkco/show-your-open-browser-tab-count-in-your-mac-menu-bar-3k96</link>
      <guid>https://dev.to/greatworkco/show-your-open-browser-tab-count-in-your-mac-menu-bar-3k96</guid>
      <description>&lt;p&gt;&lt;em&gt;This guide first appeared on &lt;a href="https://multimeterapp.com/guides/browser-tab-count" rel="noopener noreferrer"&gt;multimeterapp.com&lt;/a&gt;. Multimeter is a Mac menu bar app that turns any number into a live meter.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Tabs pile up quietly: a few windows in, nobody knows whether it’s 20 or 120. A Script meter asks Safari or Google Chrome with AppleScript how many tabs are open across all its windows and puts the number next to the clock, so you see it climb.&lt;/p&gt;

&lt;h2&gt;
  
  
  What you need
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;a href="https://multimeterapp.com/get-the-app" rel="noopener noreferrer"&gt;Multimeter&lt;/a&gt; for macOS 13 or later, $19.99 once.&lt;/li&gt;
&lt;li&gt;The direct download version of Multimeter from this site (the Script source isn’t in the Mac App Store version).&lt;/li&gt;
&lt;li&gt;Multimeter 1.1.1 or later. Earlier versions can’t ask macOS for permission to talk to other apps. Choose &lt;strong&gt;Check for Updates…&lt;/strong&gt; from any meter’s menu.&lt;/li&gt;
&lt;li&gt;Safari, Google Chrome, or both.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Set it up
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Pick the command for your browser. For Safari, this counts the tabs in every Safari window and prints a placeholder when Safari isn’t running:
&lt;/li&gt;
&lt;/ol&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;   osascript -e &lt;span class="s1"&gt;'if application "Safari" is running then'&lt;/span&gt; -e &lt;span class="s1"&gt;'tell application "Safari" to return count tabs of windows'&lt;/span&gt; -e &lt;span class="s1"&gt;'else'&lt;/span&gt; -e &lt;span class="s1"&gt;'return "{\"value\": 0, \"formatted\": \"closed\"}"'&lt;/span&gt; -e &lt;span class="s1"&gt;'end if'&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;ol&gt;
&lt;li&gt;For Google Chrome, the same line with Chrome’s name. It counts tabs in every Chrome window, Incognito windows included:
&lt;/li&gt;
&lt;/ol&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;   osascript -e &lt;span class="s1"&gt;'if application "Google Chrome" is running then'&lt;/span&gt; -e &lt;span class="s1"&gt;'tell application "Google Chrome" to return count tabs of windows'&lt;/span&gt; -e &lt;span class="s1"&gt;'else'&lt;/span&gt; -e &lt;span class="s1"&gt;'return "{\"value\": 0, \"formatted\": \"closed\"}"'&lt;/span&gt; -e &lt;span class="s1"&gt;'end if'&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;ol&gt;
&lt;li&gt;In Multimeter, choose &lt;strong&gt;Add Meter…&lt;/strong&gt; from any meter’s menu, pick &lt;strong&gt;Script&lt;/strong&gt; and paste the command. Use both commands in two meters if you use both browsers, or the combined command under Variations for one meter.&lt;/li&gt;
&lt;li&gt;Set &lt;strong&gt;Refresh every&lt;/strong&gt; to 1 minute, or 30 seconds if you want it to follow you closely. Counting takes the browser well under a second, even with 100 tabs open.&lt;/li&gt;
&lt;li&gt;Set the label to &lt;code&gt;SAFARI&lt;/code&gt;, &lt;code&gt;CHROME&lt;/code&gt; or &lt;code&gt;TABS&lt;/code&gt;, the range to 0 to 100, and, if you’d like a nudge to tidy up, alerts to High values with a warning at 30 and critical at 60. Those numbers are only a suggestion; pick the levels where you’d want to close some.&lt;/li&gt;
&lt;li&gt;With the browser open, wait for the first refresh. macOS asks once for each browser: &lt;strong&gt;“Multimeter” wants access to control “Google Chrome”. Allowing control will provide access to documents and data in “Google Chrome”, and to perform actions within that app.&lt;/strong&gt; (or the same with “Safari”). Below it is Multimeter’s reason: “Multimeter runs your script meters, and scripts that read other apps (like Things, Reminders, Mail or Music) need your permission to control them.” Click &lt;strong&gt;Allow&lt;/strong&gt;. Until you do, the meter shows a dash; the number appears at the next refresh, or right away with &lt;strong&gt;Refresh Now&lt;/strong&gt; in the meter’s menu.&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Multimeter settings
&lt;/h2&gt;

&lt;p&gt;In &lt;strong&gt;Customize This Meter…&lt;/strong&gt; (from the meter’s menu):&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Settings for this meter&lt;/em&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Setting&lt;/th&gt;
&lt;th&gt;Value&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Source&lt;/td&gt;
&lt;td&gt;Script&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Command&lt;/td&gt;
&lt;td&gt;the &lt;code&gt;osascript&lt;/code&gt; line for your browser&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Refresh every&lt;/td&gt;
&lt;td&gt;1 minute or 30 seconds&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Range&lt;/td&gt;
&lt;td&gt;0 to 100&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Alerts&lt;/td&gt;
&lt;td&gt;High values: warning 30, critical 60 (optional)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Label&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;SAFARI&lt;/code&gt;, &lt;code&gt;CHROME&lt;/code&gt; or &lt;code&gt;TABS&lt;/code&gt;
&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Face&lt;/td&gt;
&lt;td&gt;Label &amp;amp; Value or LCD Readout&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Width&lt;/td&gt;
&lt;td&gt;120 pt for Label &amp;amp; Value, 130 pt for the combined meter&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;strong&gt;Tested:&lt;/strong&gt; Set up on October 31, 2026 in Multimeter 1.1.1 on macOS 27 with Safari 27.0 and Google Chrome 156, three Label &amp;amp; Value meters (Safari, Chrome and the combined command) refreshing every 30 seconds, alerts at 30 and 60. macOS asked once for Safari and once for Chrome, each with Multimeter’s reason; after Allow the menu bar showed &lt;code&gt;SAFARI 24&lt;/code&gt;, &lt;code&gt;CHROME 41&lt;/code&gt; in yellow and &lt;code&gt;TABS 24+41&lt;/code&gt; in red. With 100 tabs across four windows in each browser, each single-browser command took 0.4 to 1 second when run in Terminal, and the combined command counted all 200 in 0.5 to 0.7 seconds, far inside the 10 second limit. Chrome’s “Allow JavaScript from Apple Events” stayed off; counting tabs doesn’t need it. With Chrome quit the meters showed &lt;code&gt;CHROME closed&lt;/code&gt; and &lt;code&gt;TABS 24+0&lt;/code&gt;; with Safari open but no windows, &lt;code&gt;SAFARI 0&lt;/code&gt;; with both quit, &lt;code&gt;closed&lt;/code&gt; on all three, and neither browser was relaunched by later refreshes.&lt;/p&gt;

&lt;p&gt;Script meters need the direct download version of Multimeter from this site, which is the one on sale now. How commands run, the PATH they see and the 10 second limit are in the &lt;a href="https://multimeterapp.com/guides/script-meters" rel="noopener noreferrer"&gt;Script meters guide&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Limits
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;The browser must be open. That’s on purpose: asking a closed browser would launch it.&lt;/li&gt;
&lt;li&gt;Safari and Google Chrome only. Other browsers have different AppleScript support, or none.&lt;/li&gt;
&lt;li&gt;Chrome Beta, Dev and Canary are separate apps; replace &lt;code&gt;Google Chrome&lt;/code&gt; with the app’s exact name, for example &lt;code&gt;Google Chrome Beta&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;The combined command names both browsers, so it needs both installed. When we ran it on a Mac without Chrome, macOS asked “Where is Google Chrome?”. If you only have one, use its own command.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Variations
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Both browsers in one meter
&lt;/h3&gt;

&lt;p&gt;This adds up Safari and Chrome. It sends JSON: &lt;code&gt;value&lt;/code&gt; is the total, which the range and alerts use, and &lt;code&gt;formatted&lt;/code&gt; is the text shown, Safari’s count plus Chrome’s, like &lt;code&gt;24+41&lt;/code&gt;. A browser that isn’t running counts as 0, and the meter says “closed” only when both are.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Command&lt;/strong&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight applescript"&gt;&lt;code&gt;&lt;span class="n"&gt;osascript&lt;/span&gt; &lt;span class="o"&gt;-&lt;/span&gt;&lt;span class="n"&gt;e&lt;/span&gt; &lt;span class="c1"&gt;'set s to 0' -e 'set c to 0' -e 'set n to 0' -e 'if application "Safari" is running then' -e 'tell application "Safari" to set s to count tabs of windows' -e 'set n to n + 1' -e 'end if' -e 'if application "Google Chrome" is running then' -e 'tell application "Google Chrome" to set c to count tabs of windows' -e 'set n to n + 1' -e 'end if' -e 'if n is 0 then return "{\"value\": 0, \"formatted\": \"closed\"}"' -e 'return "{\"value\": " &amp;amp; (s + c) &amp;amp; ", \"formatted\": \"" &amp;amp; s &amp;amp; "+" &amp;amp; c &amp;amp; "\"}"'&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  Count windows instead
&lt;/h3&gt;

&lt;p&gt;Replace &lt;code&gt;count tabs of windows&lt;/code&gt; with &lt;code&gt;count windows&lt;/code&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Troubleshooting
&lt;/h2&gt;

&lt;p&gt;&lt;em&gt;What the meter says, why, and what to do&lt;/em&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Symptom&lt;/th&gt;
&lt;th&gt;Cause&lt;/th&gt;
&lt;th&gt;Fix&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;The meter says closed&lt;/td&gt;
&lt;td&gt;The browser isn’t running. The command skips it when it’s closed so it never launches it.&lt;/td&gt;
&lt;td&gt;Open the browser. The number shows at the next refresh.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;The meter shows 0 with the browser open&lt;/td&gt;
&lt;td&gt;The browser is running with no windows open, so there are no tabs to count.&lt;/td&gt;
&lt;td&gt;Nothing to fix: open a window and the number comes back at the next refresh.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;The meter shows a dash while Safari or Google Chrome is open&lt;/td&gt;
&lt;td&gt;Multimeter isn’t allowed to control Safari or Google Chrome: you clicked Don’t Allow, or the question is still waiting on screen (the script stops after 10 seconds while it waits).&lt;/td&gt;
&lt;td&gt;Answer the prompt if it’s open. If you turned it down, open System Settings, &lt;strong&gt;Privacy &amp;amp; Security&lt;/strong&gt;, &lt;strong&gt;Automation&lt;/strong&gt;, click the arrow next to Multimeter and turn on Safari or Google Chrome. The next refresh shows the number.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Works in Terminal, nothing in the meter&lt;/td&gt;
&lt;td&gt;Terminal and Multimeter each need their own permission. Allowing Terminal to control Safari or Google Chrome doesn’t allow Multimeter.&lt;/td&gt;
&lt;td&gt;Look for Multimeter, not Terminal, under System Settings, Privacy &amp;amp; Security, Automation, and turn on Safari or Google Chrome there.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;The script didn’t print a number&lt;/td&gt;
&lt;td&gt;The command failed or printed an error message instead.&lt;/td&gt;
&lt;td&gt;Run the exact command in Terminal. If it works there, check for tools that live outside the meter’s PATH and use their full path (&lt;code&gt;which tool&lt;/code&gt;).&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;The script took longer than 10 seconds&lt;/td&gt;
&lt;td&gt;A slow network call. Multimeter stops scripts after 10 seconds.&lt;/td&gt;
&lt;td&gt;Keep the &lt;code&gt;-m&lt;/code&gt; timeout on curl, and refresh less often.&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Scripts need the version from multimeterapp.com&lt;/td&gt;
&lt;td&gt;This copy of Multimeter came from the Mac App Store, which can’t run scripts.&lt;/td&gt;
&lt;td&gt;Use the direct download version, or run the command from cron or launchd and point a &lt;a href="https://multimeterapp.com/guides/file-meters" rel="noopener noreferrer"&gt;File meter&lt;/a&gt; at its output.&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  Questions
&lt;/h2&gt;

&lt;p&gt;Script meters need the direct download version from multimeterapp.com, because the App Store sandbox doesn’t let apps run your commands. Web Address, File and Sent Value meters work in every version.&lt;/p&gt;

&lt;p&gt;No. The command asks the browser for one number, how many tabs it has, and nothing else. macOS’s prompt describes the permission in general terms (“access to documents and data”) because it covers any script you might write; the ones on this page read only the count.&lt;/p&gt;

&lt;p&gt;No. That setting lets scripts run JavaScript inside pages. Counting tabs doesn’t touch pages, and in our test it worked with the setting off, which is Chrome’s default.&lt;/p&gt;

&lt;p&gt;It depends on what they’re doing. A busy tab uses the processor, and Chrome’s Task manager names it (&lt;a href="https://multimeterapp.com/guides/google-chrome-helper-high-cpu" rel="noopener noreferrer"&gt;Why is Google Chrome Helper using so much CPU?&lt;/a&gt;); open tabs also take memory (&lt;a href="https://multimeterapp.com/guides/why-is-my-mac-slow" rel="noopener noreferrer"&gt;Why is my Mac so slow?&lt;/a&gt;). This meter counts tabs. It doesn’t measure how much they use.&lt;/p&gt;

&lt;h3&gt;
  
  
  Related
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/top-process" rel="noopener noreferrer"&gt;Show the busiest process in your Mac menu bar&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/mail-unread-count" rel="noopener noreferrer"&gt;Show your unread Mail count in your Mac menu bar&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/google-chrome-helper-high-cpu" rel="noopener noreferrer"&gt;Why is Google Chrome Helper using so much CPU on my Mac?&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/script-meters" rel="noopener noreferrer"&gt;Show the output of a shell command in your menu bar&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides/productivity" rel="noopener noreferrer"&gt;All mail and tasks recipes&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://multimeterapp.com/guides" rel="noopener noreferrer"&gt;Every guide&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>macos</category>
      <category>applescript</category>
      <category>productivity</category>
      <category>browser</category>
    </item>
    <item>
      <title>Why customer emails to your Google Group bounce, and how to let outside senders in</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Fri, 13 Nov 2026 07:59:06 +0000</pubDate>
      <link>https://dev.to/greatworkco/why-customer-emails-to-your-google-group-bounce-and-how-to-let-outside-senders-in-3m1c</link>
      <guid>https://dev.to/greatworkco/why-customer-emails-to-your-google-group-bounce-and-how-to-let-outside-senders-in-3m1c</guid>
      <description>&lt;p&gt;A lot of small teams make support@ or hello@ a Google Group instead of a mailbox. It's free with Workspace and everyone in the group gets a copy. The catch is that a new group only talks to your own organization, so the first customer who writes in gets a bounce or the message sits in moderation.&lt;/p&gt;

&lt;p&gt;Here is what to change, in order.&lt;/p&gt;

&lt;h3&gt;
  
  
  1. Let groups accept outside mail (Admin console)
&lt;/h3&gt;

&lt;p&gt;In the Google Admin console, go to Apps &amp;gt; Google Workspace &amp;gt; Groups for Business &amp;gt; Sharing settings. Two options there are off by default:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;"Group owners can allow incoming email from outside the organization"&lt;/li&gt;
&lt;li&gt;"Group owners can allow external members"&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Check the first one so customers can write in. Check the second only if you want an address outside your domain to be a member (more on that below). Save.&lt;/p&gt;

&lt;h3&gt;
  
  
  2. Open the group itself
&lt;/h3&gt;

&lt;p&gt;In Directory &amp;gt; Groups, open the group, click Access Settings and Edit. Set Who can post to include External. A setting of "Entire organization" leaves outside senders out, and their mail is rejected or held.&lt;/p&gt;

&lt;h3&gt;
  
  
  3. Check moderation and spam handling
&lt;/h3&gt;

&lt;p&gt;In the group's settings, look at Posting policies. If Message moderation holds messages from outside senders, customers' emails wait in a queue nobody watches. Spam message handling can also reject or hold messages, and those never reach members.&lt;/p&gt;

&lt;h3&gt;
  
  
  4. Know what strict DMARC does
&lt;/h3&gt;

&lt;p&gt;When the sender's domain has a strict DMARC policy (p=reject or p=quarantine), Google delivers the message as "Sender via Group", with the group's address as the sender. Before answering one of these, check who the reply is addressed to, so it goes to the customer and not back to the group.&lt;/p&gt;

&lt;h3&gt;
  
  
  5. Sending a copy somewhere outside the group
&lt;/h3&gt;

&lt;p&gt;If you want every email to support@ to also reach an outside address (a ticketing tool, a forwarding service, a personal address), add that address as a member. Since Google's 2026 change, only groups that allow external members can have an outside member. In Google Groups, open Members &amp;gt; Add members and leave "Directly add members" on. An invited address has to accept before it joins; a directly added one is a member right away.&lt;/p&gt;

&lt;p&gt;Two things to know before you do this:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Allowing external members adds an unsubscribe footer to messages sent to them.&lt;/li&gt;
&lt;li&gt;Turning the group back to internal permanently removes its external members.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The other route is a Gmail routing rule on the group with "Add more recipients", set by an admin. Rule changes can take up to 24 hours.&lt;/p&gt;

&lt;h3&gt;
  
  
  If your team answers in Slack
&lt;/h3&gt;

&lt;p&gt;We make RelayMail, which uses step 5: you add its forwarding address to the group, each email to support@ becomes a thread in a Slack channel, and a reply in the thread goes back to the customer as email. The other members still get their copies. It starts at $15 a month per inbox, with no per-seat pricing and a 14-day trial. The full walkthrough: &lt;a href="https://relaymail.app/forward-email-to-slack/google-groups" rel="noopener noreferrer"&gt;Forward a Google Group to a Slack channel&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  Quick checklist
&lt;/h3&gt;

&lt;ol&gt;
&lt;li&gt;Admin console: allow incoming email from outside the organization.&lt;/li&gt;
&lt;li&gt;Group: Who can post includes External.&lt;/li&gt;
&lt;li&gt;Moderation isn't holding outside senders.&lt;/li&gt;
&lt;li&gt;Send a test from an outside account and watch it arrive.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Sources: &lt;a href="https://knowledge.workspace.google.com/admin/groups/set-organization-wide-policies-for-using-groups?hl=en" rel="noopener noreferrer"&gt;Set organization-wide policies for using groups&lt;/a&gt;, &lt;a href="https://knowledge.workspace.google.com/admin/groups/update-group-details?hl=en" rel="noopener noreferrer"&gt;Update group details&lt;/a&gt;, &lt;a href="https://knowledge.workspace.google.com/admin/groups/changes-to-internal-and-external-classifications-in-google-groups?hl=en" rel="noopener noreferrer"&gt;Internal and external classifications in Google Groups&lt;/a&gt;, &lt;a href="https://support.google.com/groups/answer/2465464?hl=en" rel="noopener noreferrer"&gt;Add people to your group&lt;/a&gt;, &lt;a href="https://knowledge.workspace.google.com/admin/support/troubleshooting/fix-common-issues-with-group-settings?hl=en" rel="noopener noreferrer"&gt;Fix common issues with group settings&lt;/a&gt;, &lt;a href="https://support.google.com/mail/answer/1311182?hl=en" rel="noopener noreferrer"&gt;Extra info next to sender's name&lt;/a&gt;, &lt;a href="https://knowledge.workspace.google.com/admin/gmail/advanced/add-gmail-routing-settings?hl=en" rel="noopener noreferrer"&gt;Add Gmail routing settings&lt;/a&gt;.&lt;/p&gt;

</description>
      <category>googleworkspace</category>
      <category>email</category>
      <category>gsuite</category>
      <category>slack</category>
    </item>
    <item>
      <title>Gmail is ending POP and "Send as" in January 2027: what to do about support@</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Fri, 13 Nov 2026 00:28:40 +0000</pubDate>
      <link>https://dev.to/greatworkco/gmail-is-ending-pop-and-send-as-in-january-2027-what-to-do-about-support-143o</link>
      <guid>https://dev.to/greatworkco/gmail-is-ending-pop-and-send-as-in-january-2027-what-to-do-about-support-143o</guid>
      <description>&lt;p&gt;Plenty of small teams run a shared address like support@ out of one Gmail account. Gmail fetches the mail from the host over POP ("Check mail from other accounts"), and someone answers with "Send mail as". Google says both of those go away for third-party accounts in January 2027.&lt;/p&gt;

&lt;h3&gt;
  
  
  What is changing
&lt;/h3&gt;

&lt;p&gt;From Google's help pages:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;POP fetching:&lt;/strong&gt; "Gmail will no longer fetch emails from third-party accounts into your Gmail account with POP."&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Send as:&lt;/strong&gt; "You will no longer be able to send mail from third-party email accounts in Gmail on the web or in the Gmail mobile apps."&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Gmailify:&lt;/strong&gt; it takes no new users after the first quarter of 2026, and existing users can keep it until January 2027.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Google calls Q3 and Q4 2026 the transition period. In January 2027, "Gmail will fully remove 'Send as' for third-party accounts, Gmailify, and POP."&lt;/p&gt;

&lt;h3&gt;
  
  
  Who is not affected
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Addresses hosted by Google. Google says you can keep using Send as for Google-hosted identities, so a Workspace address or alias is fine.&lt;/li&gt;
&lt;li&gt;Apps that connect to Gmail itself over POP or IMAP. Google says those protocols stay supported.&lt;/li&gt;
&lt;li&gt;Mail already imported. It stays in Gmail.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;To check, open Settings &amp;gt; See all settings &amp;gt; Accounts and Import in Gmail. If "Check mail from other accounts" lists your support address, or you send as an address hosted outside Google, you need a new plan.&lt;/p&gt;

&lt;h3&gt;
  
  
  Your options
&lt;/h3&gt;

&lt;p&gt;&lt;strong&gt;Receiving.&lt;/strong&gt; Google's suggestion is automatic forwarding at the provider that hosts the address, so new mail arrives without POP.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Sending.&lt;/strong&gt; Google suggests sending from the provider's own site, or using a desktop client connected over IMAP/SMTP.&lt;/p&gt;

&lt;p&gt;That covers one person. A shared address has a second problem: forwarding to one Gmail account means one person holds the conversation. If several people answer support@, this is a good time to decide where those replies should live: a shared mailbox at your host, a help desk, or the chat tool your team already uses.&lt;/p&gt;

&lt;h3&gt;
  
  
  If your team works in Slack
&lt;/h3&gt;

&lt;p&gt;We make RelayMail, which is one option for this case. You forward support@ to a Slack channel, each email becomes a thread, and a reply in the thread goes out as a normal email. It takes a forwarding rule and no mailbox login. It starts at $15 a month per inbox, with no per-seat pricing and a 14-day trial, and it isn't a help desk, so it has no assignment or SLA tracking.&lt;/p&gt;

&lt;p&gt;We wrote up the steps here: &lt;a href="https://relaymail.app/forward-email-to-slack/gmail-pop-send-as-ending" rel="noopener noreferrer"&gt;Gmail POP and Send as are ending: forward support@ to Slack instead&lt;/a&gt;&lt;/p&gt;

&lt;h3&gt;
  
  
  Before January
&lt;/h3&gt;

&lt;ol&gt;
&lt;li&gt;Check Accounts and Import in Gmail for third-party addresses.&lt;/li&gt;
&lt;li&gt;Turn on forwarding at the host and confirm it with a test email.&lt;/li&gt;
&lt;li&gt;Decide how replies will go out, and test that too.&lt;/li&gt;
&lt;li&gt;Remove the POP account from Gmail once mail arrives in the new place.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Sources: &lt;a href="https://support.google.com/mail/answer/17101213?hl=en" rel="noopener noreferrer"&gt;Changes to third-party email account support in Gmail&lt;/a&gt;, &lt;a href="https://support.google.com/mail/answer/16604719?hl=en" rel="noopener noreferrer"&gt;Upcoming changes to Gmailify and POP&lt;/a&gt;.&lt;/p&gt;

</description>
      <category>email</category>
      <category>gmail</category>
      <category>productivity</category>
      <category>slack</category>
    </item>
    <item>
      <title>Why does NextResponse.redirect send users to localhost behind nginx?</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Thu, 12 Nov 2026 22:11:52 +0000</pubDate>
      <link>https://dev.to/greatworkco/why-does-nextresponseredirect-send-users-to-localhost-behind-nginx-57aj</link>
      <guid>https://dev.to/greatworkco/why-does-nextresponseredirect-send-users-to-localhost-behind-nginx-57aj</guid>
      <description>&lt;p&gt;&lt;em&gt;Originally published at &lt;a href="https://runheed.com/guides/nextjs-redirect-localhost-behind-proxy" rel="noopener noreferrer"&gt;runheed.com&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Under next start, a route handler's request.url is built from the address Next itself listens on, not from the Host header your proxy forwards. It does take the scheme from X-Forwarded-Proto, so behind nginx it reads &lt;a href="https://localhost:3000/" rel="noopener noreferrer"&gt;https://localhost:3000/&lt;/a&gt;... and any redirect built with new URL(path, request.url) sends the user there. Build absolute URLs from a public origin you configure, send a relative Location header, or read the forwarded host when you trust the proxy that sets it.&lt;/p&gt;

&lt;h2&gt;
  
  
  What happens
&lt;/h2&gt;

&lt;p&gt;We hit this in our own app: a form handler redirected with NextResponse.redirect(new URL(path, request.url)), and in production the browser landed on &lt;a href="https://localhost:3030" rel="noopener noreferrer"&gt;https://localhost:3030&lt;/a&gt;. nginx was passing the right Host header. We reproduced it in a fresh Next.js 16.4.5 app under next build and next start, with this route:&lt;/p&gt;

&lt;p&gt;&lt;code&gt;app/u/route.ts&lt;/code&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nc"&gt;NextRequest&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="nc"&gt;NextResponse&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="k"&gt;from&lt;/span&gt; &lt;span class="s1"&gt;'next/server'&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="k"&gt;export&lt;/span&gt; &lt;span class="kd"&gt;function&lt;/span&gt; &lt;span class="nc"&gt;GET&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nc"&gt;NextRequest&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nc"&gt;Response&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;json&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="n"&gt;url&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;url&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="n"&gt;host&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;headers&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;get&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'host'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
    &lt;span class="n"&gt;redirectTo&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nc"&gt;NextResponse&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;redirect&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="k"&gt;new&lt;/span&gt; &lt;span class="no"&gt;URL&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'/x'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;url&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;headers&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;get&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'location'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Request headers&lt;/th&gt;
&lt;th&gt;request.url&lt;/th&gt;
&lt;th&gt;Redirect Location&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Host: example.com, X-Forwarded-Proto: https&lt;/td&gt;
&lt;td&gt;&lt;a href="https://localhost:44582/u" rel="noopener noreferrer"&gt;https://localhost:44582/u&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;&lt;a href="https://localhost:44582/x" rel="noopener noreferrer"&gt;https://localhost:44582/x&lt;/a&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Host: example.com&lt;/td&gt;
&lt;td&gt;&lt;a href="http://localhost:44582/u" rel="noopener noreferrer"&gt;http://localhost:44582/u&lt;/a&gt;&lt;/td&gt;
&lt;td&gt;&lt;a href="http://localhost:44582/x" rel="noopener noreferrer"&gt;http://localhost:44582/x&lt;/a&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;The scheme follows X-Forwarded-Proto, but the host and port are the server's own, whatever Host header arrives. Local development hides it, because there the browser really is talking to localhost.&lt;/p&gt;

&lt;h2&gt;
  
  
  Fix 1: build URLs from a public origin you configure
&lt;/h2&gt;

&lt;p&gt;The most reliable fix. Set the site's public address once and use it for every absolute URL: redirects, emails, OAuth callbacks.&lt;/p&gt;

&lt;p&gt;&lt;code&gt;lib/origin.ts&lt;/code&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="k"&gt;export&lt;/span&gt; &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;publicOrigin&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="nb"&gt;process&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;env&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="no"&gt;PUBLIC_ORIGIN&lt;/span&gt; &lt;span class="o"&gt;??&lt;/span&gt; &lt;span class="s1"&gt;'http://localhost:3000'&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="c1"&gt;// in a route handler&lt;/span&gt;
&lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="nc"&gt;NextResponse&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;redirect&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="sb"&gt;`${publicOrigin()}/dashboard`&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="mi"&gt;303&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Fix 2: send a relative Location
&lt;/h2&gt;

&lt;p&gt;A redirect doesn't need an absolute URL: browsers resolve a relative Location against the page they asked for. NextResponse.redirect wants an absolute URL, so return a plain Response. In our test this came back as location: /dashboard.&lt;/p&gt;

&lt;p&gt;&lt;code&gt;app/api/resolve/route.ts&lt;/code&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="k"&gt;export&lt;/span&gt; &lt;span class="kd"&gt;function&lt;/span&gt; &lt;span class="nc"&gt;POST&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="k"&gt;new&lt;/span&gt; &lt;span class="nc"&gt;Response&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="kc"&gt;null&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;status&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="mi"&gt;303&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;headers&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="nc"&gt;Location&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s1"&gt;'/dashboard'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Fix 3: read the forwarded host, only behind a proxy you control
&lt;/h2&gt;

&lt;p&gt;The headers are there: in our test, reading them gave &lt;a href="https://example.com/x" rel="noopener noreferrer"&gt;https://example.com/x&lt;/a&gt;. But anyone can send these headers to a server that is reachable directly, and an attacker who controls the host in your redirects or password-reset links has a phishing link with your name on it. Use this only when the app is reachable solely through a proxy that sets the headers itself.&lt;/p&gt;

&lt;p&gt;&lt;code&gt;lib/request-origin.ts&lt;/code&gt;&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight typescript"&gt;&lt;code&gt;&lt;span class="k"&gt;export&lt;/span&gt; &lt;span class="kd"&gt;function&lt;/span&gt; &lt;span class="nf"&gt;requestOrigin&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nc"&gt;Request&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kt"&gt;string&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;host&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;headers&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;get&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'x-forwarded-host'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;??&lt;/span&gt; &lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;headers&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;get&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'host'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;proto&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;headers&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;get&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'x-forwarded-proto'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;??&lt;/span&gt; &lt;span class="s1"&gt;'http'&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="sb"&gt;`${proto}://${host}`&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Find the other places it bites
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Search for new URL(..., request.url) and req.nextUrl in anything that ends up in a Location header, an email or a callback URL.&lt;/li&gt;
&lt;li&gt;Reading searchParams from request.url is fine: the path and query are right, only the origin is wrong.&lt;/li&gt;
&lt;li&gt;Test behind the real proxy, or curl the server with a Host header that isn't localhost and check the Location you get back.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Does this happen in next dev?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;It does the same thing, but you won't notice: in development the browser is on localhost, so a localhost redirect works.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Is it the same on Vercel?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;We tested next start behind our own nginx, not Vercel. Check your own deployment with a request whose Host is your real domain.&lt;/p&gt;

&lt;h2&gt;
  
  
  Related
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://runheed.com/guides/node-error-handlers-next-js-traps" rel="noopener noreferrer"&gt;Four ways a Node error handler breaks in Next.js&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://runheed.com/for/nextjs" rel="noopener noreferrer"&gt;Error tracking for Next.js&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>nextjs</category>
      <category>nginx</category>
      <category>webdev</category>
      <category>javascript</category>
    </item>
    <item>
      <title>How we run dozens of coding agent sessions on one Taskhold board</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Thu, 12 Nov 2026 21:02:34 +0000</pubDate>
      <link>https://dev.to/greatworkco/how-we-run-dozens-of-coding-agent-sessions-on-one-taskhold-board-cop</link>
      <guid>https://dev.to/greatworkco/how-we-run-dozens-of-coding-agent-sessions-on-one-taskhold-board-cop</guid>
      <description>&lt;p&gt;&lt;em&gt;Originally published at &lt;a href="https://taskhold.com/guides/how-we-run-coding-agents-on-taskhold" rel="noopener noreferrer"&gt;taskhold.com&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Short answer: Great Work is one person and a lot of coding agent sessions. Every session files its work on one Taskhold board: in our first eight days that was 715 tasks, split into 2,331 steps, across 49 projects, with 55 running at once at one point. One rule in the global CLAUDE.md makes every agent do it, and the live board replaces scrolling through terminals.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;We build Taskhold, and we run our own studio on it. Great Work is one person plus coding agents working on dozens of products at the same time: web apps, Mac apps, add-ons and the sites that sell them. Since 2026-10-28 every Claude Code session on that machine reports into one Taskhold account. The numbers below are from that account, read from our own database on 2026-11-04.&lt;/p&gt;

&lt;p&gt;This is our setup, not a benchmark. If you run one agent in one repo, Claude Code's built-in task list is enough; see &lt;a href="https://taskhold.com/vs/claude-code-tasks" rel="noopener noreferrer"&gt;Claude Code tasks vs Taskhold&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  The numbers after eight days
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;Count&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Tasks filed (all by agents)&lt;/td&gt;
&lt;td&gt;715&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Steps those tasks were split into&lt;/td&gt;
&lt;td&gt;2,331&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Tasks finished&lt;/td&gt;
&lt;td&gt;442&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Projects (one per repo folder)&lt;/td&gt;
&lt;td&gt;49&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Distinct agent labels (client plus repo)&lt;/td&gt;
&lt;td&gt;44&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Tasks in progress at the time we counted&lt;/td&gt;
&lt;td&gt;55&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Times an agent stopped to ask a question&lt;/td&gt;
&lt;td&gt;28&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Our own account, 2026-10-28 to 2026-11-04. No customer data.&lt;/p&gt;

&lt;p&gt;Every task was filed by an agent, none by hand: when Jake asks for something in a session, the agent puts it on the board before it starts.&lt;/p&gt;

&lt;h2&gt;
  
  
  The one rule every session follows
&lt;/h2&gt;

&lt;p&gt;A short section in the global ~/.claude/CLAUDE.md applies to every session in every repo (the pattern is in &lt;a href="https://taskhold.com/guides/claude-md-task-workflow" rel="noopener noreferrer"&gt;the CLAUDE.md task workflow&lt;/a&gt;). The agent files each request with steps, completes steps as it goes, blocks with one question when it needs a decision, and finishes with a summary of what changed.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight markdown"&gt;&lt;code&gt;&lt;span class="gu"&gt;# tasks (Taskhold)
&lt;/span&gt;Use Taskhold for all work. I watch the live board instead of scrolling chat.

&lt;span class="k"&gt;- &lt;/span&gt;always pass project: "&amp;lt;repo folder name&amp;gt;"
&lt;span class="k"&gt;- &lt;/span&gt;every request goes on the list: taskhold_begin with title, project, steps and
  agent: "Claude Code in &amp;lt;repo&amp;gt;"
&lt;span class="k"&gt;- &lt;/span&gt;each step done: taskhold_complete with start_next: true
&lt;span class="k"&gt;- &lt;/span&gt;long step: taskhold_progress every few minutes, with a plain-sentence message
&lt;span class="k"&gt;- &lt;/span&gt;waiting on me: taskhold_block with one clear question
&lt;span class="k"&gt;- &lt;/span&gt;done: taskhold_complete on the task with what changed and how you checked it
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;A shortened version of our own rule. The full workflow snippet is on the setup page for each agent, for example &lt;a href="https://taskhold.com/for/claude-code" rel="noopener noreferrer"&gt;Claude Code&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;Projects are matched by name, so an agent working in a new repo creates its project on first use. Nobody sets anything up per repo.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the live board replaced
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Switching between terminals to see which session was still working. The &lt;a href="https://taskhold.com/demo/live" rel="noopener noreferrer"&gt;live board&lt;/a&gt; shows every running task with its progress bar, the step it is on and its last message.&lt;/li&gt;
&lt;li&gt;Questions lost in scrollback. A blocked task moves to the top under Needs you with the agent's question and suggested answers, and gets answered from a phone.&lt;/li&gt;
&lt;li&gt;Asking an agent what it did. The finished task carries a summary of what changed and how it was checked.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  What still goes wrong
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Progress is self-reported. An agent that says 80% may be stuck; the board shows when each one last reported, so a quiet session stands out.&lt;/li&gt;
&lt;li&gt;Sessions end without closing their task. Some of the 55 in-progress tasks belonged to sessions that stopped mid-way; the morning summary lists agents that went quiet so they get closed or picked up.&lt;/li&gt;
&lt;li&gt;Steps vary in size. One agent plans three steps where another plans ten, so a percentage is a rough guide, not a clock.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Do the agents need to be told to use Taskhold every time?
&lt;/h3&gt;

&lt;p&gt;No. The rule lives in the global CLAUDE.md, so every new session reads it before the first request. Per-repo setup is not needed because projects are matched by folder name.&lt;/p&gt;

&lt;h3&gt;
  
  
  Can I see the board?
&lt;/h3&gt;

&lt;p&gt;The public demo at /demo/live uses the same board components with sample tasks. Our real board has private project names, so we publish counts only.&lt;/p&gt;

&lt;h2&gt;
  
  
  Related
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/guides/monitor-coding-agent-progress" rel="noopener noreferrer"&gt;How to see what Claude Code is doing, live&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/guides/run-multiple-coding-agents-on-one-task-list" rel="noopener noreferrer"&gt;Run several coding agents in parallel on one list&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/guides/claude-md-task-workflow" rel="noopener noreferrer"&gt;A CLAUDE.md and AGENTS.md task workflow&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://taskhold.com/vs/claude-code-tasks" rel="noopener noreferrer"&gt;Taskhold vs Claude Code's built-in task list&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  About Taskhold
&lt;/h2&gt;

&lt;p&gt;Taskhold is a hosted task list shared by a person and their AI coding agents over MCP. Agents break tasks into steps and report progress, and the person watches a live board. $12/month or $120/year per person, 7-day trial with a card, no free plan. Made by Great Work LLC.&lt;/p&gt;

&lt;p&gt;To try it, add the MCP server &lt;a href="https://taskhold.com/api/mcp" rel="noopener noreferrer"&gt;https://taskhold.com/api/mcp&lt;/a&gt; to your agent. It connects with no sign-up, and the first time the agent uses Taskhold it gives you a link to start the trial.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Pricing: &lt;a href="https://taskhold.com/#pricing" rel="noopener noreferrer"&gt;https://taskhold.com/#pricing&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>ai</category>
      <category>productivity</category>
      <category>devtools</category>
      <category>showdev</category>
    </item>
    <item>
      <title>How to give Claude a phone (Claude Code, Desktop and claude.ai)</title>
      <dc:creator>Great Work</dc:creator>
      <pubDate>Thu, 12 Nov 2026 03:30:43 +0000</pubDate>
      <link>https://dev.to/greatworkco/how-to-give-claude-a-phone-claude-code-desktop-and-claudeai-5b3a</link>
      <guid>https://dev.to/greatworkco/how-to-give-claude-a-phone-claude-code-desktop-and-claudeai-5b3a</guid>
      <description>&lt;p&gt;Claude can't dial a phone on its own. Add one remote MCP server and it can: book a table, check if a part is in stock, get three quotes, or sit on hold with an airline and ring you when someone picks up.&lt;/p&gt;

&lt;p&gt;I built the server, &lt;a href="https://callforme.tel" rel="noopener noreferrer"&gt;CallForMe&lt;/a&gt;. Here's the setup for each Claude app and what actually happens on a call.&lt;/p&gt;

&lt;h2&gt;
  
  
  Claude Code
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;claude mcp add --transport http --scope user callforme https://callforme.tel/mcp
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then run &lt;code&gt;/mcp&lt;/code&gt;, pick callforme and choose Authenticate. A browser tab opens and closes by itself. There's no API key, form or card at this point.&lt;/p&gt;

&lt;h2&gt;
  
  
  Claude Desktop and claude.ai
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Open Customize, then Connectors.&lt;/li&gt;
&lt;li&gt;Choose Add, then Add custom connector.&lt;/li&gt;
&lt;li&gt;Name it CallForMe and paste &lt;code&gt;https://callforme.tel/mcp&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Click Continue, Add, then Connect.&lt;/li&gt;
&lt;li&gt;Set the tools to Always allow, so Claude doesn't ask every time it checks on a call.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Desktop and claude.ai share connectors, so one setup covers both, plus the mobile apps on the same account.&lt;/p&gt;

&lt;h2&gt;
  
  
  What happens when you ask for a call
&lt;/h2&gt;

&lt;p&gt;Ask something like:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Call Nonna's on Elm Street and book a table for 4 on Friday between 7 and 8:30. Ask for a booth. The reservation is under Sam.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;Claude then:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Calls &lt;code&gt;find_business&lt;/code&gt; for the number and whether they're open.&lt;/li&gt;
&lt;li&gt;Calls &lt;code&gt;plan_call&lt;/code&gt;, a free preview with the opening line and the questions the restaurant is likely to ask.&lt;/li&gt;
&lt;li&gt;Calls &lt;code&gt;place_call&lt;/code&gt;. The first real call returns a one-time setup link where you add $5. After that, calls just go.&lt;/li&gt;
&lt;li&gt;Polls &lt;code&gt;get_call&lt;/code&gt; until the call ends and tells you the result, with confirmation numbers and prices copied exactly.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;If the business asks something Claude didn't pass along, the question comes back to Claude during the call. Claude answers from the conversation, and the answer goes straight into the live call.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why Claude is good at this
&lt;/h2&gt;

&lt;p&gt;Claude already has the context: the dates, the car model, the order number. That makes a better call than any form, and it means an unexpected question from the business doesn't end the call.&lt;/p&gt;

&lt;h2&gt;
  
  
  Try it free first
&lt;/h2&gt;

&lt;p&gt;There are five fictional demo businesses that really pick up, free with no setup. The quickest:&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;Call Maple Street Pizza at (469) 770-7412 and book a table for 2 tonight at 7 under Alex. Ask for the confirmation number.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;p&gt;Real calls are $1 if someone answers (first 2 minutes included), then 40¢ a minute, and nothing if nobody picks up. Calls go to businesses in the US and Canada.&lt;/p&gt;

&lt;p&gt;The full guide, with tips for batches in Claude Code, is at &lt;a href="https://callforme.tel/guides/give-claude-a-phone" rel="noopener noreferrer"&gt;callforme.tel/guides/give-claude-a-phone&lt;/a&gt;.&lt;/p&gt;

</description>
      <category>claude</category>
      <category>mcp</category>
      <category>ai</category>
      <category>tutorial</category>
    </item>
  </channel>
</rss>
