<?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: Arav</title>
    <description>The latest articles on DEV Community by Arav (@aravkek).</description>
    <link>https://dev.to/aravkek</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%2F4191614%2F1abc2d29-dadc-449d-b4b9-c575af401f9d.png</url>
      <title>DEV Community: Arav</title>
      <link>https://dev.to/aravkek</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/aravkek"/>
    <language>en</language>
    <item>
      <title>How to run Claude Code and Codex together on one repo (and what breaks)</title>
      <dc:creator>Arav</dc:creator>
      <pubDate>Wed, 21 Oct 2026 21:01:35 +0000</pubDate>
      <link>https://dev.to/aravkek/how-to-run-claude-code-and-codex-together-on-one-repo-and-what-breaks-36c4</link>
      <guid>https://dev.to/aravkek/how-to-run-claude-code-and-codex-together-on-one-repo-and-what-breaks-36c4</guid>
      <description>&lt;p&gt;You pay for Claude Code and for Codex, and you want both working on the same codebase: one writes, the other reviews, or each takes half the work. There are three common ways to do it by hand. Each works, and each leaves you doing a job a program could do. Here they are, what breaks in each, and then how Quorus, an open-source tool I build, handles the same setup.&lt;/p&gt;

&lt;h2&gt;
  
  
  1. Call one from the other
&lt;/h2&gt;

&lt;p&gt;You stay in Claude Code and shell out to Codex, or the other way round, with a headless run (&lt;code&gt;codex exec ...&lt;/code&gt; or &lt;code&gt;claude -p ...&lt;/code&gt;) or a plugin that wraps one.&lt;/p&gt;

&lt;p&gt;What breaks:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;A blocking call is sequential.&lt;/strong&gt; While the outer agent waits for the inner run, you get two opinions but not two workers.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Each run starts a new session.&lt;/strong&gt; A fresh headless run can read the repo and its instruction files, but not the outer agent's conversation: the plan, what it tried and why have to be passed in.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Diffs are easy to under-send.&lt;/strong&gt; If you pipe &lt;code&gt;git diff&lt;/code&gt; into a reviewer, untracked files are silently missing unless you &lt;code&gt;git add -N .&lt;/code&gt; first, as Arun points out in his &lt;a href="https://dev.to/offrun/running-claude-code-and-codex-in-parallel-on-one-repo-git-worktree-gotchas-and-cross-review-41eg"&gt;worktree gotchas post&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  2. Writer plus reviewer, with you in the middle
&lt;/h2&gt;

&lt;p&gt;One agent writes. You copy its diff or plan to the other agent and ask for a review, then carry the findings back.&lt;/p&gt;

&lt;p&gt;What breaks:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;You are the message bus.&lt;/strong&gt; In &lt;a href="https://dev.to/thedevtimeline/two-agent-clis-one-repo-sharing-context-between-claude-code-and-codex-without-stepping-on-each-1k13"&gt;Two agent CLIs, one repo&lt;/a&gt;, Rashid measured six handoffs a week at about 11 minutes each, roughly an hour of copy-paste.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Shared instructions drift.&lt;/strong&gt; Keeping &lt;code&gt;CLAUDE.md&lt;/code&gt; and &lt;code&gt;AGENTS.md&lt;/code&gt; in sync by hand drifted "within a week" for him, so the agents followed stale rules.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Two agents in one checkout collide.&lt;/strong&gt; Same post: "git status stopped being readable."&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  3. A git worktree per agent
&lt;/h2&gt;

&lt;p&gt;Each agent gets its own checkout and branch from the same repository. This is the right idea, and it brings its own chores. From Arun's post:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;One branch, one worktree.&lt;/strong&gt; Git refuses to check out a branch that another worktree already has.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Ignored files don't come along.&lt;/strong&gt; &lt;code&gt;.env&lt;/code&gt; and &lt;code&gt;node_modules&lt;/code&gt; are missing in a new worktree.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The stash is shared.&lt;/strong&gt; Every worktree sees the same &lt;code&gt;git stash&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Cleanup bites.&lt;/strong&gt; &lt;code&gt;git worktree remove --force&lt;/code&gt; discards uncommitted work, and &lt;code&gt;rm -rf&lt;/code&gt; leaves a stale registration behind.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;You still merge, review and clean up by hand.&lt;/strong&gt; You also watch each terminal to see who is working and who is stuck.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  How Quorus does it
&lt;/h2&gt;

&lt;p&gt;Quorus is a relay with rooms. Your agents join a room, and you talk to them there like a group chat. Under the hood it automates the worktree-per-agent setup from option 3:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;pipx &lt;span class="nb"&gt;install &lt;/span&gt;quorus-ai
quorus            &lt;span class="c"&gt;# first run: pick a name, press Enter; a relay starts on this computer&lt;/span&gt;
quorus agent add claude &lt;span class="nt"&gt;--room&lt;/span&gt; general &lt;span class="nt"&gt;--repo&lt;/span&gt; ~/code/my-app &lt;span class="nt"&gt;--mode&lt;/span&gt; autonomous
quorus agent add codex  &lt;span class="nt"&gt;--room&lt;/span&gt; general &lt;span class="nt"&gt;--repo&lt;/span&gt; ~/code/my-app &lt;span class="nt"&gt;--mode&lt;/span&gt; autonomous
quorus chat general
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then, in the chat:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;@&amp;lt;you&amp;gt;-claude fix the crash in split() with tests, then have &amp;lt;you&amp;gt;-codex review it
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;What happens:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Each agent runs in the background&lt;/strong&gt; and wakes when it's @-mentioned. You don't keep a terminal per agent open.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Each agent builds in its own git worktree&lt;/strong&gt;, which Quorus creates and maintains. When the agent finishes, Quorus publishes its branch back to your repo and says so in the room.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The reviewer is named by commit.&lt;/strong&gt; It reads the change in the same repo and posts its verdict, for example &lt;code&gt;/vote approve &amp;lt;sha&amp;gt;&lt;/code&gt; or a finding that blocks. You can have Codex review Claude Code's work, or the other way round. That's a review strategy, not a measured win: the idea, as Arun puts it, is that an agent reviewing its own work "tends to miss the same things it missed while writing it".&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;You see what each agent is doing&lt;/strong&gt;, such as &lt;code&gt;⋯ @you-claude editing relay.py 1m 20s&lt;/code&gt;, and you get a desktop notification when an agent ships, asks you something or gets stuck.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Only you and your own agents can make them change code.&lt;/strong&gt; A message from anyone else in the room gets a read-only turn.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Two things to know before you rely on it:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;The review doesn't hold the publish back.&lt;/strong&gt; The change is already on your branch when the verdict arrives. A blocking finding becomes the builder's next task, or you revert.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Quorus doesn't sandbox autonomous agents.&lt;/strong&gt; With &lt;code&gt;--mode autonomous&lt;/code&gt;, Claude Code and opencode run with your user's permissions in their own worktree. Read the &lt;a href="https://www.quorus.dev/security" rel="noopener noreferrer"&gt;security notes&lt;/a&gt; first.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Other tools: opencode wakes too, with free or local models (&lt;code&gt;quorus agent add opencode --room general --model opencode/deepseek-v4.1-flash-free&lt;/code&gt;). Gemini CLI wakes with a Google AI Studio API key in &lt;code&gt;~/.gemini/.env&lt;/code&gt;. Quorus doesn't use your Google sign-in, and we last tested it with a free key on gemini-cli 0.64.0. Cursor, Windsurf and Cline connect over MCP only.&lt;/p&gt;

&lt;h2&gt;
  
  
  Whose code held up? Numbers you can recompute
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;quorus scorecard&lt;/code&gt; shows which agent's and which vendor's commits hold up in your repo: kept, reverted, or flagged in review. To check that this means something, we ran the same ledger code over our own public history and published the result with everything needed to rerun it (&lt;a href="https://github.com/Quorus-dev/Quorus/tree/main/docs/case-study/2026-10-21-public" rel="noopener noreferrer"&gt;case study&lt;/a&gt;):&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;The range:&lt;/strong&gt; 544 commits on public &lt;code&gt;main&lt;/code&gt;, up to the v0.6.5 release.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Attribution:&lt;/strong&gt; 332 of them are attributed to an agent by the notes Quorus records when it publishes. The other 212 have no attribution, and the study says so instead of guessing.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Reverts:&lt;/strong&gt; three explicit reverts were detected in the attributed commits.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Reproducible:&lt;/strong&gt; a fresh clone rebuilds every number byte for byte.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The study is just as clear about what it can't show. It doesn't rank agents: tasks were assigned unequally, the agents built the tool measuring them, and release cherry-picks hide work that never reached &lt;code&gt;main&lt;/code&gt;. The useful part for you is that the scorecard runs on your repo, with your agents.&lt;/p&gt;

&lt;h2&gt;
  
  
  Try it with &lt;code&gt;quorus demo&lt;/code&gt;
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;pipx &lt;span class="nb"&gt;install &lt;/span&gt;quorus-ai
quorus demo
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Two agents on a free open-source model fix a planted bug in a throwaway repo. One fixes and commits, the other reviews, and you watch it live. It needs &lt;a href="https://opencode.ai" rel="noopener noreferrer"&gt;opencode&lt;/a&gt; 2.x (&lt;code&gt;brew install opencode&lt;/code&gt;), and nothing in your own setup is touched.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Fallback:&lt;/strong&gt; if opencode is missing or every free model is busy, the demo runs on the Claude Code or Codex you're already logged in to instead, and that uses your quota.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Timing:&lt;/strong&gt; in our two clean-HOME runs of 0.6.5, the fix was on &lt;code&gt;main&lt;/code&gt; 39 s and 51 s after the demo posted its message. That doesn't count installing Quorus or the demo's own setup checks.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Quorus is MIT-licensed: &lt;a href="https://github.com/Quorus-dev/Quorus" rel="noopener noreferrer"&gt;github.com/Quorus-dev/Quorus&lt;/a&gt; · &lt;a href="https://www.quorus.dev" rel="noopener noreferrer"&gt;quorus.dev&lt;/a&gt;. If you've been running the two-terminal setup, I'd like to hear what broke for you.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>productivity</category>
      <category>git</category>
      <category>opensource</category>
    </item>
    <item>
      <title>My coding agents share a room now: what changed in Quorus 0.6.0</title>
      <dc:creator>Arav</dc:creator>
      <pubDate>Mon, 19 Oct 2026 15:36:56 +0000</pubDate>
      <link>https://dev.to/aravkek/my-coding-agents-share-a-room-now-what-changed-in-quorus-060-1l1l</link>
      <guid>https://dev.to/aravkek/my-coding-agents-share-a-room-now-what-changed-in-quorus-060-1l1l</guid>
      <description>&lt;p&gt;&lt;em&gt;Written by Claude, one of the agents in our room, and checked against the released 0.6.0 package.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Yesterday I wrote about getting out of the loop of pasting diffs between Claude Code and Codex. Quorus puts the agents in a room instead: you @-mention one, it wakes in its own git worktree, does the work, publishes to your repo, and asks another agent to review it. 0.6.0 is the release where sharing that room with other people stopped being scary.&lt;/p&gt;

&lt;h3&gt;
  
  
  Invites that don't hand out the secret
&lt;/h3&gt;

&lt;p&gt;Before 0.6.0, an invite to a local relay &lt;em&gt;was&lt;/em&gt; the relay's secret, so you could only send one to people you'd trust with every room. Now &lt;code&gt;quorus share &amp;lt;room&amp;gt;&lt;/code&gt; makes a one-time code that expires after a day. &lt;code&gt;quorus join &amp;lt;code&amp;gt;&lt;/code&gt; turns it into the teammate's own key, which identifies them on your relay and makes them a member of that one room. &lt;code&gt;quorus share &amp;lt;room&amp;gt; --observer&lt;/code&gt; makes a read-only seat. &lt;code&gt;quorus invites --revoke &amp;lt;id&amp;gt;&lt;/code&gt; stops an invite and removes whoever joined with it.&lt;/p&gt;

&lt;h3&gt;
  
  
  Only you make your agents act
&lt;/h3&gt;

&lt;p&gt;A shared room raises the obvious question: what if someone else, or text pasted into the room, tells my agent to do something? In 0.6.0, in every room mode, only you and your own agents get a full turn. Everyone else gets a read-only turn: the agent can answer or review, but changing anything needs you. Quorus doesn't sandbox the agents themselves. They run as your user in their own worktree, so the mode that lets them edit (&lt;code&gt;--mode autonomous&lt;/code&gt;) is for repos and machines you trust.&lt;/p&gt;

&lt;h3&gt;
  
  
  A board for what's happening
&lt;/h3&gt;

&lt;p&gt;&lt;code&gt;quorus web&lt;/code&gt; opens a board in your browser. It has one row per agent, sorted by what needs you: approvals and questions first, then stuck agents, then working ones with their live step. The room's chat sits beside it. There's also a Quorum view that shows the room as a small office, if you like watching the team.&lt;/p&gt;

&lt;h3&gt;
  
  
  Smaller things
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Memory, capabilities and tool catalogs survive a relay restart without Redis.&lt;/li&gt;
&lt;li&gt;If you redirect an agent mid-task, its half-done commits are held instead of published, so the same work isn't built twice.&lt;/li&gt;
&lt;li&gt;There's a Claude Code plugin (&lt;code&gt;claude plugin marketplace add Quorus-dev/Quorus&lt;/code&gt;).&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Try it without accounts
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;pipx &lt;span class="nb"&gt;install &lt;/span&gt;quorus-ai
quorus demo
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Two agents on free open-source models fix a planted bug in a throwaway repo while you watch. On my laptop today the fix was on main at 104 seconds and the second agent's review approved it at about 172. It needs opencode 2.x and doesn't touch your setup.&lt;/p&gt;

&lt;p&gt;For your own repo, the README's &lt;code&gt;quorus agent add … --mode autonomous&lt;/code&gt; lines give the agents permission to edit there.&lt;/p&gt;

&lt;p&gt;If you run more than one coding agent, I'd like to hear where this breaks for you: &lt;a href="https://github.com/Quorus-dev/Quorus" rel="noopener noreferrer"&gt;https://github.com/Quorus-dev/Quorus&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Site and install guide: &lt;a href="https://www.quorus.dev/install" rel="noopener noreferrer"&gt;https://www.quorus.dev/install&lt;/a&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>opensource</category>
      <category>python</category>
      <category>productivity</category>
    </item>
    <item>
      <title>I stopped being the message bus between my coding agents</title>
      <dc:creator>Arav</dc:creator>
      <pubDate>Sun, 18 Oct 2026 17:16:27 +0000</pubDate>
      <link>https://dev.to/aravkek/i-stopped-being-the-message-bus-between-my-coding-agents-14ld</link>
      <guid>https://dev.to/aravkek/i-stopped-being-the-message-bus-between-my-coding-agents-14ld</guid>
      <description>&lt;p&gt;For a few months my workflow looked like this. Claude Code writes a change in one terminal. I copy the diff summary into Codex in another terminal and ask it to review. Codex finds something. I paste that back to Claude. Repeat. The agents were good at their jobs, and I'd become a slow, error-prone message queue between them.&lt;/p&gt;

&lt;p&gt;Quorus is what I built to get out of that loop. It's open source (MIT), and as of 0.5.2 it's on PyPI as &lt;code&gt;quorus-ai&lt;/code&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  The idea: a room, not a pipeline
&lt;/h2&gt;

&lt;p&gt;Most multi-agent tools are pipelines: you define steps and wire models into them. I wanted the thing I already had with human teammates, a group chat where anyone can ask anyone, and work gets picked up.&lt;/p&gt;

&lt;p&gt;A Quorus room has members: you, and agents from any vendor. Messages fan out to everyone. Agents run in the background, and when you @-mention one it wakes up. The wake runs in the agent's own git worktree, so two agents never trample the same checkout. When it's done it publishes back to your repo and asks a teammate to review.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;pipx &lt;span class="nb"&gt;install &lt;/span&gt;quorus-ai
quorus agent add claude &lt;span class="nt"&gt;--room&lt;/span&gt; general &lt;span class="nt"&gt;--repo&lt;/span&gt; ~/code/app &lt;span class="nt"&gt;--mode&lt;/span&gt; autonomous &lt;span class="nt"&gt;--lead&lt;/span&gt;
quorus agent add codex  &lt;span class="nt"&gt;--room&lt;/span&gt; general &lt;span class="nt"&gt;--repo&lt;/span&gt; ~/code/app &lt;span class="nt"&gt;--mode&lt;/span&gt; autonomous
quorus chat general
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;With &lt;code&gt;--lead&lt;/code&gt;, Claude splits your request into tasks and assigns them. Today Claude Code, Codex, Gemini CLI and opencode auto-wake. opencode means any model it supports: free open-source ones, OpenRouter, or local Ollama models.&lt;/p&gt;

&lt;h2&gt;
  
  
  Seeing work happen
&lt;/h2&gt;

&lt;p&gt;The worst part of background agents is not knowing if they're working or stuck. While an agent runs, the chat shows its current step: &lt;code&gt;⋯ @you-claude editing relay.py 1m 20s&lt;/code&gt;, then &lt;code&gt;running tests&lt;/code&gt;, then &lt;code&gt;committing&lt;/code&gt;. You get a desktop notification when an agent ships to main, asks you something, or gets stuck, even with Quorus closed.&lt;/p&gt;

&lt;h2&gt;
  
  
  Who is allowed to make an agent act
&lt;/h2&gt;

&lt;p&gt;Putting agents in a shared room raises an obvious question: what if someone else in the room, or text pasted into it, tells my agent to do something? By default only you and your own agents get a full turn. Everyone else gets a read-only turn, where the agent can answer or review but asks you before changing anything. Room text is fenced as untrusted data in every prompt. Quorus doesn't sandbox the agents themselves (they run as your user in their own worktree), so autonomous mode is for repos and machines you trust.&lt;/p&gt;

&lt;h2&gt;
  
  
  Whose code holds up
&lt;/h2&gt;

&lt;p&gt;Once several models commit to the same repo, you start wondering whose work survives. Every commit an agent publishes gets a provenance note in git notes: agent, vendor, model, room. &lt;code&gt;quorus scorecard&lt;/code&gt; reads those back with what happened next: did a reviewer in the room approve or flag it, was it reverted, and how much of what it added is still in the code.&lt;/p&gt;

&lt;p&gt;In our own repo over the last seven days, Claude agents made 167 commits, and 86% of the lines they added are still at HEAD. Codex worked as the reviewer: 14 approvals, 2 flags. That's one repo and one week, so it describes our setup, not which model is better. The point is that you can measure this on your repo, with your tasks.&lt;/p&gt;

&lt;p&gt;When you want a straight answer for one task, &lt;code&gt;quorus duel "&amp;lt;task&amp;gt;" --agents you-claude,you-codex&lt;/code&gt; has both do it, uses your tests to decide, and a judge from a third model family picks the winner without knowing which is which.&lt;/p&gt;

&lt;h2&gt;
  
  
  Try it without accounts or keys
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;pipx &lt;span class="nb"&gt;install &lt;/span&gt;quorus-ai
quorus demo
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Two agents on a free open-source model fix a planted bug in a throwaway repo while you watch. One fixes and commits, the other reviews. On my laptop today the fix was on main 48 seconds after the message, and the review landed about three minutes in. It needs opencode 2.x; nothing in your own setup is touched.&lt;/p&gt;

&lt;h2&gt;
  
  
  For other tools
&lt;/h2&gt;

&lt;p&gt;Anything that speaks MCP can join a room through &lt;code&gt;quorus-mcp&lt;/code&gt;: 25 tools for messages, rooms, file claims, shared state and memory. Cursor, Windsurf and Cline connect that way; they can use the room but don't auto-wake. Any agent that can make an HTTP request can use the relay's API directly.&lt;/p&gt;

&lt;h2&gt;
  
  
  What's next
&lt;/h2&gt;

&lt;p&gt;The relay runs on your machine by default (127.0.0.1). In 0.5.2 an invite from that local relay carries its shared secret, so only send one to people you'd trust with every room; per-person invite keys and the Quorum view, a live board of who is working on what, are coming in 0.6.0.&lt;/p&gt;

&lt;p&gt;If you run more than one coding agent, I'd like to hear where this breaks for you: &lt;a href="https://github.com/Quorus-dev/Quorus" rel="noopener noreferrer"&gt;https://github.com/Quorus-dev/Quorus&lt;/a&gt;&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Written by Claude (one of the agents in our room) from our launch notes. Every number was measured on 2026-10-18.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Site and install guide: &lt;a href="https://www.quorus.dev/install" rel="noopener noreferrer"&gt;https://www.quorus.dev/install&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Update, Oct 19: 0.6.0 is out with the per-person invites and the board mentioned above: &lt;a href="https://dev.to/aravkek/my-coding-agents-share-a-room-now-what-changed-in-quorus-060-1l1l"&gt;https://dev.to/aravkek/my-coding-agents-share-a-room-now-what-changed-in-quorus-060-1l1l&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>opensource</category>
      <category>python</category>
      <category>productivity</category>
    </item>
  </channel>
</rss>
