<?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: 橋本透</title>
    <description>The latest articles on DEV Community by 橋本透 (@winclawdev).</description>
    <link>https://dev.to/winclawdev</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%2F3765888%2Fadd4d683-e210-4062-96b8-51df81aae8b8.png</url>
      <title>DEV Community: 橋本透</title>
      <link>https://dev.to/winclawdev</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/winclawdev"/>
    <language>en</language>
    <item>
      <title>Biz AI's integration skill: let Claude Code, Codex or MetaCoder AI-enable your existing business system</title>
      <dc:creator>橋本透</dc:creator>
      <pubDate>Mon, 16 Nov 2026 02:21:06 +0000</pubDate>
      <link>https://dev.to/winclawdev/biz-ais-integration-skill-let-claude-code-codex-or-metacoder-ai-enable-your-existing-business-3kmm</link>
      <guid>https://dev.to/winclawdev/biz-ais-integration-skill-let-claude-code-codex-or-metacoder-ai-enable-your-existing-business-3kmm</guid>
      <description>&lt;p&gt;Meta's consumer agent &lt;strong&gt;Muse&lt;/strong&gt; works on its own on schedules and events, remembers earlier conversations, and asks on an approval card before sensitive steps. &lt;strong&gt;Biz AI&lt;/strong&gt; does the same &lt;strong&gt;inside a company's business system, with each employee's own permissions&lt;/strong&gt; — open source (Apache-2.0), not affiliated with Meta. The AI proposes; confirming, approving and amounts always stay with people.&lt;/p&gt;

&lt;p&gt;Connecting an existing system still means adding things on the business side — employees and roles, a live-session check, recipient rules, approvals, reports, a personal-data check: 13 items in total. Working all of that out by hand takes time.&lt;/p&gt;

&lt;p&gt;So Biz AI ships an &lt;strong&gt;integration skill for AI coding agents&lt;/strong&gt;. This post shows how to put it into &lt;strong&gt;Claude Code, Codex, DeepSeek Harness and MetaCoder&lt;/strong&gt; and let them do the integration.&lt;/p&gt;

&lt;blockquote&gt;
&lt;p&gt;The design is in &lt;a href="https://dev.to/winclawdev/biz-ai-meta-muse-style-proactive-ai-agents-inside-business-systems-and-why-the-ai-must-never-1l1j"&gt;an earlier post&lt;/a&gt;.&lt;/p&gt;
&lt;/blockquote&gt;

&lt;h2&gt;
  
  
  What is in the skill
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;biz-assistant-integration/
  SKILL.md                   entry point: choosing a path, how to proceed, rules, what people decide
  references/                ten procedures, read only when needed
    01-assessment.md         assessing an existing system (13 items)
    02-new-system.md         building a new system
    03-existing-system.md    connecting an existing one (Node.js, aliases, other languages, external mode)
    04-manifest.md           writing and validating the manifest
    05-business-side.md      hooks, events, recipients, approvals, proposals, reports
    06-dsh-profile.md        AI-plane profile, outbound settings, automatic tasks
    07-web-and-bff.md        business web screens and the BFF
    08-optional-features.md  manual, memory, profile, attachments, external connectors
    09-rules.md              MUST / SHOULD summary
    10-verification.md       conformance and the definition of done
  templates/
    integration-plan.md      integration plan template
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The agent reads &lt;code&gt;SKILL.md&lt;/code&gt; first and opens only the reference it needs right now. &lt;strong&gt;The repository's documents are the source of truth&lt;/strong&gt;; the skill is the route through them.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the skill makes the agent do
&lt;/h2&gt;

&lt;p&gt;It goes &lt;strong&gt;assess → plan → implement → verify&lt;/strong&gt;:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Step&lt;/th&gt;
&lt;th&gt;What&lt;/th&gt;
&lt;th&gt;How it is checked&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;1 Assess and plan&lt;/td&gt;
&lt;td&gt;read login, employees and roles, APIs, transactions, batch jobs; fill the 13 items as exists / add / alias; choose a path; write &lt;code&gt;integration-plan.md&lt;/code&gt;
&lt;/td&gt;
&lt;td&gt;a person reviews the plan&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;2 Manifest&lt;/td&gt;
&lt;td&gt;roles and delegation ceilings, events and payload allow-lists, messages, tools, playbooks, reports&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;bridge-validate manifest --strict&lt;/code&gt; exits 0&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;3 Business side&lt;/td&gt;
&lt;td&gt;hooks (&lt;code&gt;employees&lt;/code&gt;, &lt;code&gt;authenticate&lt;/code&gt;, &lt;code&gt;isSessionLive&lt;/code&gt;, &lt;code&gt;recipientsFor&lt;/code&gt;, &lt;code&gt;approvals&lt;/code&gt;, &lt;code&gt;proposalKinds&lt;/code&gt;, &lt;code&gt;reports&lt;/code&gt;, &lt;code&gt;piiScan&lt;/code&gt;); AI writes are proposals only&lt;/td&gt;
&lt;td&gt;the system's own tests&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;4 Events&lt;/td&gt;
&lt;td&gt;the business row and &lt;code&gt;emitEvent&lt;/code&gt; in the same transaction&lt;/td&gt;
&lt;td&gt;the system's own tests&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;5 Conformance&lt;/td&gt;
&lt;td&gt;52 checks against a TEST environment&lt;/td&gt;
&lt;td&gt;&lt;code&gt;52 passed, 0 known gaps, 0 skipped&lt;/code&gt;&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;6 AI plane&lt;/td&gt;
&lt;td&gt;profile, bot, outbound settings kept closed&lt;/td&gt;
&lt;td&gt;profile checks&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;7 Web and BFF&lt;/td&gt;
&lt;td&gt;screens for chat linking, notification settings, approvals, AI proposals, the AI employee's profile&lt;/td&gt;
&lt;td&gt;type checks, screens&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;8 Automatic tasks&lt;/td&gt;
&lt;td&gt;the delegation class and the playbooks&lt;/td&gt;
&lt;td&gt;profile checks&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;There are five paths: &lt;strong&gt;new&lt;/strong&gt; on server-kit / &lt;strong&gt;existing Node.js&lt;/strong&gt; with server-kit mounted / &lt;strong&gt;aliases&lt;/strong&gt; on existing APIs / &lt;strong&gt;another language&lt;/strong&gt; implementing Bridge v1 and proven by the conformance suite / &lt;strong&gt;external mode&lt;/strong&gt; without code changes (trial).&lt;/p&gt;

&lt;h3&gt;
  
  
  It does not decide for you
&lt;/h3&gt;

&lt;p&gt;The skill has a list of &lt;strong&gt;what people decide&lt;/strong&gt;. The agent stops there, lays out the options with a recommendation, and waits — for example:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;the first business flow and what the AI may read&lt;/li&gt;
&lt;li&gt;which events go to whom (and which may be sent at night)&lt;/li&gt;
&lt;li&gt;what the AI may propose and which roles may accept it&lt;/li&gt;
&lt;li&gt;whose delegation automatic tasks run under&lt;/li&gt;
&lt;li&gt;whether any closed outbound setting is opened, and why&lt;/li&gt;
&lt;li&gt;which chat and which model&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Rules it must keep
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;never modify the framework itself (&lt;code&gt;packages/*&lt;/code&gt;, &lt;code&gt;plugins/*&lt;/code&gt;, &lt;code&gt;vendor/*&lt;/code&gt;, the profile template); fill gaps with hooks and settings&lt;/li&gt;
&lt;li&gt;the AI's tools are GETs and proposals only; business write APIs return 403 to a delegation&lt;/li&gt;
&lt;li&gt;the business system decides recipients; if it can't, nothing is sent&lt;/li&gt;
&lt;li&gt;no personal data in events or messages (IDs, dates, counts, states only)&lt;/li&gt;
&lt;li&gt;no secrets in the repository, config files, logs or chat — a person enters the values&lt;/li&gt;
&lt;li&gt;fictional data, loopback and a TEST environment only&lt;/li&gt;
&lt;li&gt;never claim "it connects without changes"; the plan lists what is added&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Installing it in each tool
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/ITC-cloud-soft/biz-ai.git
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then copy the folder &lt;code&gt;biz-ai/docs/skill/biz-assistant-integration&lt;/code&gt; to where your tool looks for skills:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Tool&lt;/th&gt;
&lt;th&gt;For you (all projects)&lt;/th&gt;
&lt;th&gt;For the team (the business system's repo)&lt;/th&gt;
&lt;th&gt;How it is used&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Claude Code&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;code&gt;~/.claude/skills/&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;&lt;code&gt;&amp;lt;repo&amp;gt;/.claude/skills/&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;picked automatically when the request matches the skill's description&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Codex&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;&lt;code&gt;~/.agents/skills/&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;&lt;code&gt;&amp;lt;repo&amp;gt;/.agents/skills/&lt;/code&gt;&lt;/td&gt;
&lt;td&gt;automatically, or &lt;code&gt;$biz-assistant-integration&lt;/code&gt; / &lt;code&gt;/skills&lt;/code&gt;
&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;DeepSeek Harness&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;~/.dsh/skills/&lt;/code&gt; or &lt;code&gt;~/.agents/skills/&lt;/code&gt;
&lt;/td&gt;
&lt;td&gt;
&lt;code&gt;&amp;lt;repo&amp;gt;/.dsh/skills/&lt;/code&gt; or &lt;code&gt;.agents/skills/&lt;/code&gt;
&lt;/td&gt;
&lt;td&gt;picked when the request matches&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;MetaCoder&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;—&lt;/td&gt;
&lt;td&gt;put it in the repo&lt;/td&gt;
&lt;td&gt;ask it to read &lt;code&gt;SKILL.md&lt;/code&gt; first&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# Claude Code (team)&lt;/span&gt;
cp -r biz-ai/docs/skill/biz-assistant-integration &lt;span class="o"&gt;&amp;lt;&lt;/span&gt;repo&lt;span class="o"&gt;&amp;gt;&lt;/span&gt;/.claude/skills/

&lt;span class="c"&gt;# Codex and DeepSeek Harness both read .agents/skills — one copy serves both&lt;/span&gt;
mkdir -p &lt;span class="o"&gt;&amp;lt;&lt;/span&gt;repo&lt;span class="o"&gt;&amp;gt;&lt;/span&gt;/.agents/skills
cp -r biz-ai/docs/skill/biz-assistant-integration &lt;span class="o"&gt;&amp;lt;&lt;/span&gt;repo&lt;span class="o"&gt;&amp;gt;&lt;/span&gt;/.agents/skills/
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;strong&gt;MetaCoder&lt;/strong&gt; (&lt;a href="https://metacoder.myaiportal.net/" rel="noopener noreferrer"&gt;metacoder.myaiportal.net&lt;/a&gt;) is our AI development tool: it reads the whole codebase as a semantic knowledge graph and drives development through a 10-phase harness pipeline with machine-checked quality gates. Put the skill in the business system's repo and have it read &lt;code&gt;SKILL.md&lt;/code&gt; first; having the whole system as a graph should suit step 1, the 13-item assessment.&lt;/p&gt;

&lt;h2&gt;
  
  
  How to ask
&lt;/h2&gt;

&lt;p&gt;Open the agent in the business system's repository and ask in your own words:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;Integrate Biz Assistant Framework into this system.
この業務システムに Bizアシスタントを導入して。最初は請求の承認から。
把 biz assistant framework 导入这个既有系统，先从巡检超期提醒开始。
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Tip: &lt;strong&gt;start with one business flow&lt;/strong&gt; (invoice approval, overdue inspections, …), run it next to the existing screens, then widen.&lt;/p&gt;

&lt;h2&gt;
  
  
  "Done" is decided by commands
&lt;/h2&gt;

&lt;p&gt;The last step checks the definition of done with commands: strict manifest validation, &lt;strong&gt;conformance 52 / 52&lt;/strong&gt; (0 skipped, 0 known gaps), the system's tests, profile checks, outbound settings, type checks. Anything not checked (a real chat, a real model) must be written down as "not verified".&lt;/p&gt;

&lt;p&gt;The conformance suite calls the system from the outside over HTTP, so you can see &lt;strong&gt;whether the code the agent wrote breaks a Bridge rule&lt;/strong&gt; without reading all of it. Add an employee's name to an event payload, for instance, and the write is refused and the tests go red.&lt;/p&gt;

&lt;h2&gt;
  
  
  The best and most complete guide is on the official site
&lt;/h2&gt;

&lt;p&gt;This post is the entrance. &lt;strong&gt;The integration flow, how it works, the safety commitments and how to use the skill are explained in full on the official site.&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Official site: &lt;a href="https://biz-ai.myaiportal.net/en/" rel="noopener noreferrer"&gt;https://biz-ai.myaiportal.net/en/&lt;/a&gt;&lt;/strong&gt;

&lt;ul&gt;
&lt;li&gt;Integration and the skill: &lt;a href="https://biz-ai.myaiportal.net/en/integration.html" rel="noopener noreferrer"&gt;https://biz-ai.myaiportal.net/en/integration.html&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;li&gt;GitHub: &lt;a href="https://github.com/ITC-cloud-soft/biz-ai" rel="noopener noreferrer"&gt;https://github.com/ITC-cloud-soft/biz-ai&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Questions: &lt;a href="https://github.com/ITC-cloud-soft/biz-ai/discussions" rel="noopener noreferrer"&gt;GitHub Discussions&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>ai</category>
      <category>agents</category>
      <category>claudecode</category>
      <category>opensource</category>
    </item>
    <item>
      <title>Biz AI: Meta Muse-style proactive AI agents, inside business systems — and why the AI must never approve</title>
      <dc:creator>橋本透</dc:creator>
      <pubDate>Sun, 15 Nov 2026 02:12:14 +0000</pubDate>
      <link>https://dev.to/winclawdev/biz-ai-meta-muse-style-proactive-ai-agents-inside-business-systems-and-why-the-ai-must-never-1l1j</link>
      <guid>https://dev.to/winclawdev/biz-ai-meta-muse-style-proactive-ai-agents-inside-business-systems-and-why-the-ai-must-never-1l1j</guid>
      <description>&lt;h2&gt;
  
  
  Meta Muse, inside a company
&lt;/h2&gt;

&lt;p&gt;In September 2026 Meta launched &lt;strong&gt;Muse&lt;/strong&gt;, a personal AI agent that doesn't just chat. It works in the background on schedules and events, remembers earlier conversations, acts on your behalf, and asks on an &lt;strong&gt;approval card&lt;/strong&gt; before sensitive steps.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Biz AI does the same inside a company's business system, for each employee&lt;/strong&gt; — an inspection log, a billing system, a leave-request app. It is open source (Apache-2.0) and not affiliated with Meta.&lt;/p&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;Meta Muse (consumer)&lt;/th&gt;
&lt;th&gt;Biz AI (business systems)&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;works for&lt;/td&gt;
&lt;td&gt;one person's life&lt;/td&gt;
&lt;td&gt;each employee of a company, by role&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;acts on its own&lt;/td&gt;
&lt;td&gt;on schedules and events&lt;/td&gt;
&lt;td&gt;on business events (overdue, requested, unanswered…) and scheduled reports&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;remembers&lt;/td&gt;
&lt;td&gt;earlier conversations, memory&lt;/td&gt;
&lt;td&gt;per-employee and organization memory (personal data checked before saving)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;approvals&lt;/td&gt;
&lt;td&gt;approval card before sensitive steps&lt;/td&gt;
&lt;td&gt;approval cards + reply codes; &lt;strong&gt;the AI cannot write, only propose&lt;/strong&gt;
&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;permissions&lt;/td&gt;
&lt;td&gt;the accounts the user connected&lt;/td&gt;
&lt;td&gt;the employee's own role × a manifest; delegations issued by the business system&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;where you talk&lt;/td&gt;
&lt;td&gt;the Muse app&lt;/td&gt;
&lt;td&gt;the chat each employee already uses (Slack, LINE, Teams, …)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;source of truth&lt;/td&gt;
&lt;td&gt;each service&lt;/td&gt;
&lt;td&gt;
&lt;strong&gt;the company's business system&lt;/strong&gt; — every result is recorded there&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;A consumer agent gets more useful the more it decides for you. At work you have to be able to explain afterwards whose permission the agent read with, whom it told and who approved. So Biz AI is proactive like Muse, and &lt;strong&gt;people make every decision&lt;/strong&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it looks like
&lt;/h2&gt;

&lt;p&gt;From &lt;code&gt;pnpm demo&lt;/code&gt; (a fictional equipment-inspection system; the demo UI is Japanese, translated here):&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;manager ▶ Any overdue inspections?
  AI employee ▶ One overdue inspection (FE-002).

business system ▶ (detects an overdue inspection: inspection.overdue)
  AI employee ▶ inspector1  [on its own]
    │ [Please handle] An inspection is 3 days overdue (due 2026-11-11) …
    │ [ Open the business web ]

inspector1 ▶ (registers a repair request in the app, estimate 12,000 yen)
  AI employee ▶ manager  [on its own]
    │ [Approval request] Repair request · equipment FE-002 · estimate 12,000 yen
    │ Reply by text: "承認 321904" / "差戻 321904 &amp;lt;reason&amp;gt;"
    │ [ Approve ] [ Send back ] [ Details on the business web ]

manager ▶ (taps [Approve] on the card)
  AI employee ▶ Approved. The card is updated too.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Run it in three minutes — Node 22.19+, pnpm and Docker; no API key, no chat account:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;git clone https://github.com/ITC-cloud-soft/biz-ai.git &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="nb"&gt;cd&lt;/span&gt; biz-ai
pnpm install
pnpm demo   &lt;span class="c"&gt;# real agent runtime, scripted model, mock chat, throwaway PostgreSQL&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Guardrails in code, not in prompts
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;1. The AI reads only with the employee's own permissions.&lt;/strong&gt; The read tools it may call are the manifest's allow-list intersected with the employee's role. A delegation token on a write endpoint gets &lt;code&gt;403 delegation_not_permitted&lt;/code&gt;; the AI can only file a &lt;em&gt;proposal&lt;/em&gt; that a person accepts or rejects.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="c1"&gt;// a read the AI may call = a manifest tool; every write is people only&lt;/span&gt;
&lt;span class="n"&gt;router&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;add&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'GET'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="s1"&gt;'/api/leave-requests'&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="n"&gt;kit&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;readToolPolicy&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s1"&gt;'GET /api/leave-requests'&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="no"&gt;ROLES&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;async&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;principal&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;query&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="p"&gt;{&lt;/span&gt;
    &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;list&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="p"&gt;.&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;data&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;leave&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;values&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="nf"&gt;filter&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nf"&gt;visibleTo&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;principal&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;user&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="c1"&gt;// staff see their own only&lt;/span&gt;
    &lt;span class="k"&gt;return&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;200&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;body&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;ok&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="n"&gt;leaveRequests&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;list&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;map&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;leaveJson&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;p&gt;&lt;strong&gt;2. The business system decides who is told&lt;/strong&gt;, with the same rules as its web screens. The agent never guesses recipients.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="n"&gt;recipientsFor&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="k"&gt;async&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;event&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;event&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="kd"&gt;type&lt;/span&gt; &lt;span class="o"&gt;!==&lt;/span&gt; &lt;span class="s1"&gt;'leave.unanswered'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="kc"&gt;null&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;l&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;data&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;leave&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="n"&gt;event&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;payload&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;leaveRequestId&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="n"&gt;l&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;stale&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s1"&gt;'deleted'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;l&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;status&lt;/span&gt; &lt;span class="o"&gt;!==&lt;/span&gt; &lt;span class="s1"&gt;'pending'&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;stale&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s1"&gt;'decided'&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;   &lt;span class="c1"&gt;// already decided: tell nobody&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;users&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="p"&gt;.&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;data&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;users&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;values&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="nf"&gt;filter&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;u&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="n"&gt;u&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;active&lt;/span&gt; &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="no"&gt;MANAGERS&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;includes&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;u&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;role&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;||&lt;/span&gt; &lt;span class="n"&gt;u&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;id&lt;/span&gt; &lt;span class="o"&gt;===&lt;/span&gt; &lt;span class="n"&gt;l&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;requesterUserId&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="nf"&gt;map&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;employee&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;
  &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;users&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;dedupeKey&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="sb"&gt;`leave.unanswered:${l.id}`&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="cm"&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;p&gt;&lt;strong&gt;3. Events are written in the same transaction as the business change&lt;/strong&gt; (an outbox), so a notice is never sent for something that didn't happen, or missed for something that did.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;4. One decision closes every approval card&lt;/strong&gt;, and the requester can never approve their own request.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the business system has to add
&lt;/h2&gt;

&lt;p&gt;It doesn't connect with zero changes. The first diagnosis lists what the system adds (employees and roles, a live-session check, recipient rules, approvals, proposal kinds, reports, a personal-data check — 13 items).&lt;/p&gt;

&lt;p&gt;The smallest example, &lt;code&gt;examples/leave-requests&lt;/code&gt;, is &lt;strong&gt;one ~300-line file with no database&lt;/strong&gt;, and passes the &lt;strong&gt;52-check Bridge v1 conformance suite&lt;/strong&gt; (0 skipped, 0 known gaps). Copy it out of the repo and it runs on the npm packages alone:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;cp -r biz-ai/examples/leave-requests ./my-system &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; &lt;span class="nb"&gt;cd&lt;/span&gt; my-system
npm pkg &lt;span class="nb"&gt;set&lt;/span&gt; dependencies.@dsh-bridge/protocol=^0.1.1 dependencies.@dsh-bridge/server-kit=^0.1.1
npm install &lt;span class="o"&gt;&amp;amp;&amp;amp;&lt;/span&gt; npm &lt;span class="nb"&gt;test&lt;/span&gt;    &lt;span class="c"&gt;# 57 tests: business rules + 52 conformance checks&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The suite is the safety net while you turn it into your own business: add an employee's name to an event payload and &lt;code&gt;kit.emitEvent&lt;/code&gt; refuses it, and six tests go red — personal data is stopped before it reaches the AI side.&lt;/p&gt;

&lt;h2&gt;
  
  
  The AI side
&lt;/h2&gt;

&lt;p&gt;Nine plugins — identity linking, manifest-generated read tools, events, notifications, approvals, an assistant, memory, a persona and document reading — built as DeepSeek Harness plugins (&lt;code&gt;@dsh-bridge/*&lt;/code&gt; on npm).&lt;/p&gt;

&lt;p&gt;For an existing system, point a coding agent (Claude Code or another) at the bundled integration skill (&lt;code&gt;docs/skill/biz-assistant-integration&lt;/code&gt;): it goes diagnose → plan → implement → verify and stops to ask whenever a decision is yours. Start with &lt;strong&gt;one&lt;/strong&gt; business flow, run it next to the existing screens, then widen.&lt;/p&gt;

&lt;p&gt;Coding agents are strong enough now. The next step is AI that works with people inside a company — with people making the decisions.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Repo: &lt;a href="https://github.com/ITC-cloud-soft/biz-ai" rel="noopener noreferrer"&gt;https://github.com/ITC-cloud-soft/biz-ai&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Site: &lt;a href="https://biz-ai.myaiportal.net/en/" rel="noopener noreferrer"&gt;https://biz-ai.myaiportal.net/en/&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Discussions: &lt;a href="https://github.com/ITC-cloud-soft/biz-ai/discussions" rel="noopener noreferrer"&gt;https://github.com/ITC-cloud-soft/biz-ai/discussions&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>ai</category>
      <category>agents</category>
      <category>opensource</category>
      <category>node</category>
    </item>
    <item>
      <title>WinClaw's 18 Business Plugins: From Invoice Processing to Inventory Management — All AI-Powered</title>
      <dc:creator>橋本透</dc:creator>
      <pubDate>Mon, 02 Mar 2026 11:19:56 +0000</pubDate>
      <link>https://dev.to/winclawdev/winclaws-18-business-plugins-from-invoice-processing-to-inventory-management-all-ai-powered-20li</link>
      <guid>https://dev.to/winclawdev/winclaws-18-business-plugins-from-invoice-processing-to-inventory-management-all-ai-powered-20li</guid>
      <description>&lt;h2&gt;
  
  
  Beyond Development: AI-Powered Business Automation
&lt;/h2&gt;

&lt;p&gt;WinClaw isn't just for developers. It ships with &lt;strong&gt;18 business plugins&lt;/strong&gt; that automate everyday office tasks using AI. Each plugin works through natural conversation — just describe what you need, and the AI handles the rest.&lt;/p&gt;

&lt;h2&gt;
  
  
  The 18 Plugins at a Glance
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Document &amp;amp; Data Processing
&lt;/h3&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Plugin&lt;/th&gt;
&lt;th&gt;What It Does&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;invoice-processor&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Extract data from invoices (PDF/image), validate amounts, export to CSV/Excel&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;receipt-scanner&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Scan receipts, categorize expenses, generate expense reports&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;document-converter&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Convert between formats (PDF, Word, Excel, CSV, JSON) with layout preservation&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;data-entry-automator&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Auto-fill forms and spreadsheets from source documents&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h3&gt;
  
  
  Communication &amp;amp; Scheduling
&lt;/h3&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Plugin&lt;/th&gt;
&lt;th&gt;What It Does&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;email-drafter&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Draft professional emails from brief notes, with tone/style control&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;meeting-scheduler&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Find optimal meeting times, send invites, handle timezone conversion&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;minutes-generator&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Generate meeting minutes from audio transcripts or notes&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;translation-bridge&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Real-time document translation (JP/EN/ZH) with domain-specific terminology&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h3&gt;
  
  
  Finance &amp;amp; Accounting
&lt;/h3&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Plugin&lt;/th&gt;
&lt;th&gt;What It Does&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;budget-tracker&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Track project budgets, flag overruns, generate variance reports&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;payroll-calculator&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Calculate wages, deductions, overtime with locale-specific tax rules&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;expense-reporter&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Compile and categorize expenses, generate approval-ready reports&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h3&gt;
  
  
  Inventory &amp;amp; Operations
&lt;/h3&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Plugin&lt;/th&gt;
&lt;th&gt;What It Does&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;inventory-manager&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Track stock levels, predict reorder points, generate purchase orders&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;shipping-tracker&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Monitor shipments across carriers, alert on delays, update ETAs&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;order-processor&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Process incoming orders, validate stock, generate fulfillment tasks&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h3&gt;
  
  
  Reporting &amp;amp; Analytics
&lt;/h3&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Plugin&lt;/th&gt;
&lt;th&gt;What It Does&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;report-generator&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Create formatted business reports from raw data with charts&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;kpi-dashboard&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Build real-time KPI dashboards with drill-down capabilities&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;trend-analyzer&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Identify trends in sales/operations data with forecasting&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;compliance-checker&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;Verify documents against regulatory requirements (GDPR, SOX, etc.)&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  How Plugins Work
&lt;/h2&gt;

&lt;p&gt;Each plugin follows a simple pattern:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# Example: Process invoices&lt;/span&gt;
winclaw plugin invoice-processor &lt;span class="nt"&gt;--input&lt;/span&gt; ./invoices/ &lt;span class="nt"&gt;--output&lt;/span&gt; ./processed/

&lt;span class="c"&gt;# Example: Generate meeting minutes&lt;/span&gt;
winclaw plugin minutes-generator &lt;span class="nt"&gt;--transcript&lt;/span&gt; meeting-2024-03.txt

&lt;span class="c"&gt;# Example: Track inventory&lt;/span&gt;
winclaw plugin inventory-manager &lt;span class="nt"&gt;--action&lt;/span&gt; reorder-check
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The AI reads your input, understands the context, processes the data, and outputs structured results.&lt;/p&gt;

&lt;h2&gt;
  
  
  Real-World Example: Invoice Processing
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Drop invoice PDFs into a folder&lt;/li&gt;
&lt;li&gt;Run the invoice-processor plugin&lt;/li&gt;
&lt;li&gt;AI extracts: vendor, date, line items, amounts, tax&lt;/li&gt;
&lt;li&gt;Outputs a clean CSV ready for your accounting software&lt;/li&gt;
&lt;li&gt;Flags any discrepancies or missing information&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Getting Started
&lt;/h2&gt;

&lt;p&gt;All 18 plugins are included free with WinClaw:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/itc-ou-shigou/winclaw" rel="noopener noreferrer"&gt;GitHub&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;
&lt;a href="https://sourceforge.net/projects/winclaw/files/" rel="noopener noreferrer"&gt;SourceForge&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nb"&gt;set &lt;/span&gt;&lt;span class="nv"&gt;ANTHROPIC_BASE_URL&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;https://api.z.ai/api/anthropic
&lt;span class="nb"&gt;set &lt;/span&gt;&lt;span class="nv"&gt;ANTHROPIC_AUTH_TOKEN&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;your-api-key
&lt;span class="nb"&gt;set &lt;/span&gt;&lt;span class="nv"&gt;ANTHROPIC_MODEL&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;glm-5
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Conclusion
&lt;/h2&gt;

&lt;p&gt;WinClaw's 18 business plugins bring AI automation to everyday office tasks. From invoice processing to inventory management, each plugin saves hours of manual work. Best of all, they're completely free and open-source. Give them a try!&lt;/p&gt;

</description>
    </item>
    <item>
      <title>WinClaw AI Testing: Fully Automated Web App Testing Without Writing Test Scripts</title>
      <dc:creator>橋本透</dc:creator>
      <pubDate>Mon, 02 Mar 2026 11:10:24 +0000</pubDate>
      <link>https://dev.to/winclawdev/winclaw-ai-testing-fully-automated-web-app-testing-without-writing-test-scripts-3lbc</link>
      <guid>https://dev.to/winclawdev/winclaw-ai-testing-fully-automated-web-app-testing-without-writing-test-scripts-3lbc</guid>
      <description>&lt;h2&gt;
  
  
  The Problem with Traditional Testing
&lt;/h2&gt;

&lt;p&gt;Writing Selenium scripts, maintaining Playwright selectors, updating tests every time the UI changes... Manual test maintenance is a massive time sink for development teams.&lt;/p&gt;

&lt;p&gt;What if an AI could look at your web app and test it automatically — no scripts needed?&lt;/p&gt;

&lt;h2&gt;
  
  
  WinClaw AI Dev System Testing
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;WinClaw&lt;/strong&gt; is a free, open-source AI development tool for Windows. Its &lt;strong&gt;ai-dev-system-testing&lt;/strong&gt; skill uses Claude Code + Chrome automation to test web applications completely automatically.&lt;/p&gt;

&lt;h3&gt;
  
  
  How It Works
&lt;/h3&gt;

&lt;p&gt;The testing workflow runs in 5 phases:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Phase 1: Environment Check&lt;/strong&gt; — Verifies Chrome debug port, MCP connections, and target URLs&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Phase 2: Discovery&lt;/strong&gt; — AI crawls your app, maps all pages, identifies forms and interactive elements&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Phase 3: Frontend Testing&lt;/strong&gt; — Tests UI rendering, responsive design, form validation, navigation&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Phase 4: API Testing&lt;/strong&gt; — Discovers and tests backend endpoints (supports Swagger/OpenAPI auto-detection)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Phase 5: Iteration&lt;/strong&gt; — Re-tests failed items with different strategies, auto-fixes where possible&lt;/li&gt;
&lt;/ol&gt;

&lt;h3&gt;
  
  
  What Gets Tested Automatically
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Page rendering&lt;/strong&gt;: Screenshots + accessibility checks on every page&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Form validation&lt;/strong&gt;: Fills forms with valid/invalid data, checks error messages&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Navigation&lt;/strong&gt;: Tests all links, breadcrumbs, routing&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Responsive design&lt;/strong&gt;: Mobile/tablet/desktop viewport testing&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;API endpoints&lt;/strong&gt;: GET/POST/PUT/DELETE with schema validation&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Error handling&lt;/strong&gt;: 404 pages, network errors, edge cases&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Zero Configuration
&lt;/h3&gt;

&lt;p&gt;Just point WinClaw at your running app:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# Frontend + Backend&lt;/span&gt;
winclaw &lt;span class="nb"&gt;test&lt;/span&gt; &lt;span class="nt"&gt;--frontend&lt;/span&gt; http://localhost:3000 &lt;span class="nt"&gt;--backend&lt;/span&gt; http://localhost:3001/docs

&lt;span class="c"&gt;# Frontend only&lt;/span&gt;
winclaw &lt;span class="nb"&gt;test&lt;/span&gt; &lt;span class="nt"&gt;--frontend&lt;/span&gt; http://localhost:3000
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The AI figures out what to test, generates test scenarios, executes them, and produces a detailed report — all without a single line of test code.&lt;/p&gt;

&lt;h3&gt;
  
  
  Test Reports
&lt;/h3&gt;

&lt;p&gt;After testing completes, you get:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Pass rate&lt;/strong&gt; by category (UI, API, forms, navigation)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Screenshots&lt;/strong&gt; of every tested page&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Detailed failure logs&lt;/strong&gt; with reproduction steps&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Recommendations&lt;/strong&gt; for fixes&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Getting Started
&lt;/h2&gt;

&lt;p&gt;WinClaw is completely free:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/itc-ou-shigou/winclaw" rel="noopener noreferrer"&gt;GitHub&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://sourceforge.net/projects/winclaw/files/" rel="noopener noreferrer"&gt;SourceForge&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;LLM backend (GLM-5, free):&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nb"&gt;set &lt;/span&gt;&lt;span class="nv"&gt;ANTHROPIC_BASE_URL&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;https://api.z.ai/api/anthropic
&lt;span class="nb"&gt;set &lt;/span&gt;&lt;span class="nv"&gt;ANTHROPIC_AUTH_TOKEN&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;your-api-key
&lt;span class="nb"&gt;set &lt;/span&gt;&lt;span class="nv"&gt;ANTHROPIC_MODEL&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;glm-5
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Conclusion
&lt;/h2&gt;

&lt;p&gt;WinClaw's AI testing eliminates the need for manual test scripts. Point it at your app, let the AI discover and test everything, and get a comprehensive report. It's like having a QA engineer that never sleeps and never needs training. Try it out!&lt;/p&gt;

</description>
      <category>ai</category>
      <category>automation</category>
      <category>opensource</category>
      <category>testing</category>
    </item>
    <item>
      <title>WinClaw: Auto-Deploy to AWS, Azure &amp; Alibaba Cloud with AI-Powered Conversations</title>
      <dc:creator>橋本透</dc:creator>
      <pubDate>Mon, 02 Mar 2026 11:04:02 +0000</pubDate>
      <link>https://dev.to/winclawdev/winclaw-auto-deploy-to-aws-azure-alibaba-cloud-with-ai-powered-conversations-4pjo</link>
      <guid>https://dev.to/winclawdev/winclaw-auto-deploy-to-aws-azure-alibaba-cloud-with-ai-powered-conversations-4pjo</guid>
      <description>&lt;h2&gt;
  
  
  Introduction
&lt;/h2&gt;

&lt;p&gt;"I want to deploy to AWS, but writing CloudFormation YAML is a pain..." "Azure has too many configuration options..." Sound familiar?&lt;/p&gt;

&lt;p&gt;I had the same frustrations until I tried &lt;strong&gt;WinClaw&lt;/strong&gt;'s cloud auto-deploy skills. Just by having a conversation with AI, I got my app deployed to &lt;strong&gt;AWS, Azure, and Alibaba Cloud&lt;/strong&gt; — fully automated.&lt;/p&gt;

&lt;h2&gt;
  
  
  What is WinClaw Cloud Deploy?
&lt;/h2&gt;

&lt;p&gt;WinClaw is a free, open-source AI development tool with three cloud deployment skills:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;aws-cloud-deploy&lt;/strong&gt;: Auto-generates CloudFormation templates&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;azure-cloud-deploy&lt;/strong&gt;: Auto-generates ARM templates&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;aliyun-cloud-deploy&lt;/strong&gt;: Auto-generates ROS templates&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;All three providers share the same conversational workflow — great for multi-cloud environments.&lt;/p&gt;

&lt;h2&gt;
  
  
  The Interactive Deploy Experience
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Phase 1: Requirements Hearing
&lt;/h3&gt;

&lt;p&gt;The AI asks you natural questions:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Monthly budget&lt;/strong&gt;: "How much can you spend per month?" → e.g., $50/mo&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Expected traffic&lt;/strong&gt;: "Daily visitors?" → e.g., 1,000 PV/day&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Security needs&lt;/strong&gt;: "Handling personal data?" → Yes/No&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Availability&lt;/strong&gt;: "Acceptable downtime?" → e.g., &amp;lt; 30 min/month&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Phase 2: Architecture Proposal
&lt;/h3&gt;

&lt;p&gt;Based on your answers, the AI proposes the optimal architecture from 6 patterns:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Pattern&lt;/th&gt;
&lt;th&gt;Cost/mo&lt;/th&gt;
&lt;th&gt;Use Case&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Lite&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;$10-40&lt;/td&gt;
&lt;td&gt;Single EC2, blogs/testing&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Standard&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;$50-150&lt;/td&gt;
&lt;td&gt;EC2 + RDS, small web apps&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;HA&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;$150-300&lt;/td&gt;
&lt;td&gt;ALB + Auto Scaling + Multi-AZ&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Elastic&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;$250-600&lt;/td&gt;
&lt;td&gt;CloudFront + ElastiCache&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Serverless&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;$0-100&lt;/td&gt;
&lt;td&gt;Lambda + API Gateway&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;&lt;strong&gt;Container&lt;/strong&gt;&lt;/td&gt;
&lt;td&gt;$300+&lt;/td&gt;
&lt;td&gt;ECS/EKS microservices&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h3&gt;
  
  
  Security Tiers
&lt;/h3&gt;

&lt;p&gt;Security is also auto-selected from 3 tiers:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Basic&lt;/strong&gt; (Free): Security Groups + IAM Roles&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Standard&lt;/strong&gt; ($10-30/mo): WAF + VPC Flow Logs + Encryption&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Enterprise&lt;/strong&gt; ($200-3,500/mo): Shield Advanced + GuardDuty + Config Rules + Audit Logs&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Phase 3: Automated Deployment
&lt;/h3&gt;

&lt;p&gt;After confirming the architecture, everything runs automatically:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;Phase 3A&lt;/strong&gt; — IaC template generation (CloudFormation/ARM/ROS)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Phase 3B&lt;/strong&gt; — Code deployment&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Phase 3C&lt;/strong&gt; — Health checks + auto-fix on errors&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Phase 3D&lt;/strong&gt; — Report generation (access URL, cost estimate)&lt;/li&gt;
&lt;/ol&gt;

&lt;h2&gt;
  
  
  Same Experience on Azure &amp;amp; Alibaba Cloud
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Azure&lt;/strong&gt; generates ARM templates and selects from App Service / VM / AKS / Functions.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Alibaba Cloud&lt;/strong&gt; generates ROS templates and selects from ECS / FC / ACK. Supports China's MLPS 2.0 compliance.&lt;/p&gt;

&lt;h2&gt;
  
  
  Getting Started
&lt;/h2&gt;

&lt;p&gt;WinClaw is completely free and open-source:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;&lt;a href="https://github.com/itc-ou-shigou/winclaw" rel="noopener noreferrer"&gt;GitHub Repository&lt;/a&gt;&lt;/li&gt;
&lt;li&gt;&lt;a href="https://sourceforge.net/projects/winclaw/files/" rel="noopener noreferrer"&gt;SourceForge Downloads&lt;/a&gt;&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Use GLM-5 (free) as the LLM backend:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nb"&gt;set &lt;/span&gt;&lt;span class="nv"&gt;ANTHROPIC_BASE_URL&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;https://api.z.ai/api/anthropic
&lt;span class="nb"&gt;set &lt;/span&gt;&lt;span class="nv"&gt;ANTHROPIC_AUTH_TOKEN&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;your-api-key
&lt;span class="nb"&gt;set &lt;/span&gt;&lt;span class="nv"&gt;ANTHROPIC_MODEL&lt;/span&gt;&lt;span class="o"&gt;=&lt;/span&gt;glm-5
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Conclusion
&lt;/h2&gt;

&lt;p&gt;WinClaw's cloud deploy skills dramatically lower the barrier to infrastructure setup. Just answer questions about budget and traffic, get an optimal architecture proposal, and watch as everything deploys automatically. Once you try it, there's no going back. Give it a shot!&lt;/p&gt;

</description>
      <category>ai</category>
      <category>automation</category>
      <category>cloud</category>
      <category>devops</category>
    </item>
    <item>
      <title>OpenClaw: An Open-Source AI Assistant That Runs Natively on Windows PC — With Office, Outlook &amp; Multi-Channel Support</title>
      <dc:creator>橋本透</dc:creator>
      <pubDate>Wed, 11 Feb 2026 11:22:35 +0000</pubDate>
      <link>https://dev.to/winclawdev/openclaw-an-open-source-ai-assistant-that-runs-natively-on-windows-pc-with-office-outlook--23g7</link>
      <guid>https://dev.to/winclawdev/openclaw-an-open-source-ai-assistant-that-runs-natively-on-windows-pc-with-office-outlook--23g7</guid>
      <description>&lt;h2&gt;
  
  
  TL;DR
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;OpenClaw (WinClaw)&lt;/strong&gt; is an open-source &lt;strong&gt;personal AI assistant / multi-channel AI gateway&lt;/strong&gt; that connects to WhatsApp, Telegram, Discord, Signal, Slack, iMessage, Microsoft Teams, and more.&lt;/p&gt;

&lt;p&gt;The biggest highlight: &lt;strong&gt;it runs natively on Windows PC&lt;/strong&gt; with a bundled EXE installer (Node.js 22 included, zero prerequisites). It also supports &lt;strong&gt;macOS, Linux, and Raspberry Pi&lt;/strong&gt;.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;GitHub:&lt;/strong&gt; &lt;a href="https://github.com/itc-ou-shigou/winclaw" rel="noopener noreferrer"&gt;https://github.com/itc-ou-shigou/winclaw&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;npm:&lt;/strong&gt; &lt;code&gt;npm install -g openclaw&lt;/code&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Docs:&lt;/strong&gt; &lt;a href="https://docs.openclaw.ai" rel="noopener noreferrer"&gt;https://docs.openclaw.ai&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;License:&lt;/strong&gt; MIT&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Why OpenClaw?
&lt;/h2&gt;

&lt;p&gt;Every time you want to ask AI something, you open ChatGPT's web interface. What if your AI assistant lived inside the messaging apps you already use every day?&lt;/p&gt;

&lt;p&gt;With OpenClaw, you just talk to your AI through WhatsApp, Discord, or Telegram. It responds using Claude, GPT, Gemini, or even local LLMs via Ollama. No browser tabs needed.&lt;/p&gt;

&lt;h2&gt;
  
  
  Native Windows PC Support
&lt;/h2&gt;

&lt;p&gt;Unlike most AI tools that focus on cloud or Linux, OpenClaw is built with &lt;strong&gt;Windows-first&lt;/strong&gt; thinking:&lt;/p&gt;

&lt;h3&gt;
  
  
  EXE Installer (Recommended)
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Bundled Node.js 22 runtime, zero prerequisites&lt;/li&gt;
&lt;li&gt;Download from &lt;a href="https://github.com/itc-ou-shigou/winclaw/releases" rel="noopener noreferrer"&gt;GitHub Releases&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;Auto-start via Windows Task Scheduler on login&lt;/li&gt;
&lt;li&gt;No admin rights required&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Other Install Methods
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight powershell"&gt;&lt;code&gt;&lt;span class="c"&gt;# PowerShell one-liner&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="n"&gt;irm&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;https://raw.githubusercontent.com/itc-ou-shigou/winclaw/main/install.ps1&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="o"&gt;|&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="n"&gt;iex&lt;/span&gt;&lt;span class="w"&gt;

&lt;/span&gt;&lt;span class="c"&gt;# winget&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="n"&gt;winget&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;install&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;WinClaw.WinClaw&lt;/span&gt;&lt;span class="w"&gt;

&lt;/span&gt;&lt;span class="c"&gt;# npm&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="n"&gt;npm&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;install&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;-g&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;openclaw&lt;/span&gt;&lt;span class="err"&gt;@&lt;/span&gt;&lt;span class="nx"&gt;latest&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  macOS / Linux
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npm &lt;span class="nb"&gt;install&lt;/span&gt; &lt;span class="nt"&gt;-g&lt;/span&gt; openclaw@latest
openclaw onboard &lt;span class="nt"&gt;--install-daemon&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;Platform&lt;/th&gt;
&lt;th&gt;Daemon Method&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Windows&lt;/td&gt;
&lt;td&gt;schtasks (Task Scheduler)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;macOS&lt;/td&gt;
&lt;td&gt;launchd user service&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Linux&lt;/td&gt;
&lt;td&gt;systemd user service&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  Skill System: Office &amp;amp; Outlook Automation
&lt;/h2&gt;

&lt;p&gt;OpenClaw's &lt;strong&gt;skill system&lt;/strong&gt; dramatically extends your AI assistant's capabilities:&lt;/p&gt;

&lt;h3&gt;
  
  
  Built-in Skills
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Windows Office&lt;/strong&gt; — Create/edit Word, Excel, PowerPoint (python-docx, openpyxl, python-pptx)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Outlook Integration&lt;/strong&gt; — Send/read/manage emails (COM Automation)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Windows Explorer&lt;/strong&gt; — File search, bulk rename, compression, clipboard&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Windows System&lt;/strong&gt; — Services, processes, registry, event logs&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Browser Automation&lt;/strong&gt; — Control Chrome pages via extension&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Coding Agent&lt;/strong&gt; — Codex CLI, Claude Code programming support&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Custom Skills
&lt;/h3&gt;

&lt;p&gt;Skills are &lt;code&gt;SKILL.md&lt;/code&gt; + scripts. Create and share freely. Community skills on &lt;a href="https://clawhub.com" rel="noopener noreferrer"&gt;ClawHub&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Supported AI Models
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Anthropic Claude&lt;/strong&gt; — Opus 4.6, Sonnet, Haiku (OAuth + API key)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;OpenAI&lt;/strong&gt; — GPT-4o, o1, o3, Codex&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Google Gemini&lt;/strong&gt; — 2.5 Pro, Flash&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Local LLMs&lt;/strong&gt; — via Ollama&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Supported Messaging Channels
&lt;/h2&gt;

&lt;p&gt;WhatsApp | Telegram | Discord | Signal | Slack | Google Chat | iMessage | Microsoft Teams | Matrix | Zalo | WebChat&lt;/p&gt;

&lt;p&gt;Multiple channels simultaneously.&lt;/p&gt;

&lt;h2&gt;
  
  
  Agent Capabilities
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;File operations, web search, browser automation&lt;/li&gt;
&lt;li&gt;Shell command execution&lt;/li&gt;
&lt;li&gt;Scheduled tasks (cron)&lt;/li&gt;
&lt;li&gt;Camera/screen capture (node pairing)&lt;/li&gt;
&lt;li&gt;Long-term memory system&lt;/li&gt;
&lt;li&gt;Sub-agent spawning&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Quick Start
&lt;/h2&gt;

&lt;h3&gt;
  
  
  Windows
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight powershell"&gt;&lt;code&gt;&lt;span class="n"&gt;winclaw&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;onboard&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nt"&gt;--flow&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;quickstart&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="n"&gt;winclaw&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;daemon&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;install&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;span class="n"&gt;winclaw&lt;/span&gt;&lt;span class="w"&gt; &lt;/span&gt;&lt;span class="nx"&gt;dashboard&lt;/span&gt;&lt;span class="w"&gt;
&lt;/span&gt;&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h3&gt;
  
  
  macOS / Linux
&lt;/h3&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;openclaw onboard &lt;span class="nt"&gt;--install-daemon&lt;/span&gt;
openclaw dashboard
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;h2&gt;
  
  
  Summary
&lt;/h2&gt;

&lt;p&gt;OpenClaw aims to be &lt;strong&gt;the AI assistant that runs on your own PC (especially Windows)&lt;/strong&gt;. With built-in skills for Office and Outlook, it directly supports your daily workflow.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;GitHub:&lt;/strong&gt; &lt;a href="https://github.com/itc-ou-shigou/winclaw" rel="noopener noreferrer"&gt;https://github.com/itc-ou-shigou/winclaw&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;npm:&lt;/strong&gt; &lt;a href="https://www.npmjs.com/package/openclaw" rel="noopener noreferrer"&gt;https://www.npmjs.com/package/openclaw&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Docs:&lt;/strong&gt; &lt;a href="https://docs.openclaw.ai" rel="noopener noreferrer"&gt;https://docs.openclaw.ai&lt;/a&gt;
&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Discord:&lt;/strong&gt; &lt;a href="https://discord.com/invite/clawd" rel="noopener noreferrer"&gt;https://discord.com/invite/clawd&lt;/a&gt;
&lt;/li&gt;
&lt;/ul&gt;

</description>
      <category>selfhosted</category>
    </item>
    <item>
      <title>WinClaw: I Built an Open-Source AI Assistant That Connects to All Your Messaging Apps</title>
      <dc:creator>橋本透</dc:creator>
      <pubDate>Wed, 11 Feb 2026 08:50:41 +0000</pubDate>
      <link>https://dev.to/winclawdev/winclaw-i-built-an-open-source-ai-assistant-that-connects-to-all-your-messaging-apps-a0</link>
      <guid>https://dev.to/winclawdev/winclaw-i-built-an-open-source-ai-assistant-that-connects-to-all-your-messaging-apps-a0</guid>
      <description>&lt;h2&gt;
  
  
  TL;DR
&lt;/h2&gt;

&lt;p&gt;I built &lt;strong&gt;WinClaw&lt;/strong&gt;, an open-source personal AI assistant that runs locally on your machine and connects to 11+ messaging channels. Think of it as a self-hosted AI gateway that lets you talk to Claude, ChatGPT, or any AI model through WhatsApp, Telegram, Slack, Discord, Signal, and more.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;GitHub:&lt;/strong&gt; &lt;a href="https://github.com/itc-ou-shigou/winclaw" rel="noopener noreferrer"&gt;github.com/itc-ou-shigou/winclaw&lt;/a&gt;&lt;/p&gt;




&lt;h2&gt;
  
  
  The Problem
&lt;/h2&gt;

&lt;p&gt;AI assistants are trapped in their own apps. Want to use Claude? Open claude.ai. ChatGPT? Open chatgpt.com. But what if you want your AI assistant available everywhere you already chat — WhatsApp with friends, Slack at work, Discord in communities, Signal for privacy?&lt;/p&gt;

&lt;h2&gt;
  
  
  The Solution: WinClaw
&lt;/h2&gt;

&lt;p&gt;WinClaw is a personal AI gateway that:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Connects to 11+ channels&lt;/strong&gt; — WhatsApp, Telegram, Slack, Discord, Signal, iMessage, Microsoft Teams, Matrix, Google Chat, Zalo, and WebChat&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Runs on any OS&lt;/strong&gt; — Windows (standalone EXE installer), macOS (launchd), Linux (systemd)&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Supports multiple AI models&lt;/strong&gt; — Claude, OpenAI, or any OpenAI-compatible provider with automatic failover&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Keeps data local&lt;/strong&gt; — Everything stays on your machine. No cloud dependency.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;5-minute setup&lt;/strong&gt; — Interactive onboarding wizard guides you through everything&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Key Features
&lt;/h2&gt;

&lt;h3&gt;
  
  
  🧠 Memory System
&lt;/h3&gt;

&lt;p&gt;Your assistant remembers context across sessions with daily notes and long-term memory files.&lt;/p&gt;

&lt;h3&gt;
  
  
  🔧 Extensible Skills
&lt;/h3&gt;

&lt;p&gt;Browse 100+ skills at &lt;a href="https://clawhub.com" rel="noopener noreferrer"&gt;clawhub.com&lt;/a&gt; or create your own. Skills let your assistant do everything from managing emails to controlling smart home devices.&lt;/p&gt;

&lt;h3&gt;
  
  
  🌐 Browser Automation
&lt;/h3&gt;

&lt;p&gt;Chrome Browser Relay lets your assistant browse the web, fill forms, and automate tasks in your browser.&lt;/p&gt;

&lt;h3&gt;
  
  
  📱 Device Pairing
&lt;/h3&gt;

&lt;p&gt;Pair phones, cameras, and IoT devices. Your assistant can snap photos, check screens, and control hardware.&lt;/p&gt;

&lt;h3&gt;
  
  
  ⏰ Proactive Scheduling
&lt;/h3&gt;

&lt;p&gt;Cron jobs and heartbeat monitoring let your assistant check emails, calendars, and notifications without being asked.&lt;/p&gt;

&lt;h3&gt;
  
  
  🔄 Multi-Model Failover
&lt;/h3&gt;

&lt;p&gt;Configure multiple AI providers with automatic failover. If one goes down, your assistant seamlessly switches.&lt;/p&gt;

&lt;h2&gt;
  
  
  Installation
&lt;/h2&gt;



&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="c"&gt;# Windows (recommended)&lt;/span&gt;
winget &lt;span class="nb"&gt;install &lt;/span&gt;WinClaw.WinClaw

&lt;span class="c"&gt;# macOS / Linux&lt;/span&gt;
npm &lt;span class="nb"&gt;install&lt;/span&gt; &lt;span class="nt"&gt;-g&lt;/span&gt; openclaw@latest
openclaw onboard &lt;span class="nt"&gt;--install-daemon&lt;/span&gt;

&lt;span class="c"&gt;# Docker&lt;/span&gt;
docker pull openclaw/openclaw:latest
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;On Windows, you can also download the standalone EXE installer from &lt;a href="https://github.com/itc-ou-shigou/winclaw/releases" rel="noopener noreferrer"&gt;GitHub Releases&lt;/a&gt; — zero prerequisites, bundled Node.js runtime.&lt;/p&gt;

&lt;h2&gt;
  
  
  Tech Stack
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Runtime:&lt;/strong&gt; Node.js 22 LTS&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Recommended Model:&lt;/strong&gt; Anthropic Claude Opus 4.6&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;License:&lt;/strong&gt; MIT&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Platforms:&lt;/strong&gt; Windows 10+, macOS, Linux&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Install methods:&lt;/strong&gt; EXE, npm, winget, Docker, Nix&lt;/li&gt;
&lt;/ul&gt;

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

&lt;p&gt;I'm actively working on:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;More skills and integrations&lt;/li&gt;
&lt;li&gt;Improved multi-language support&lt;/li&gt;
&lt;li&gt;Community skill marketplace&lt;/li&gt;
&lt;li&gt;Better documentation and tutorials&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Try It Out
&lt;/h2&gt;

&lt;p&gt;👉 &lt;strong&gt;GitHub:&lt;/strong&gt; &lt;a href="https://github.com/itc-ou-shigou/winclaw" rel="noopener noreferrer"&gt;github.com/itc-ou-shigou/winclaw&lt;/a&gt;&lt;br&gt;
📚 &lt;strong&gt;Docs:&lt;/strong&gt; &lt;a href="https://docs.openclaw.ai" rel="noopener noreferrer"&gt;docs.openclaw.ai&lt;/a&gt;&lt;br&gt;
🧩 &lt;strong&gt;Skills:&lt;/strong&gt; &lt;a href="https://clawhub.com" rel="noopener noreferrer"&gt;clawhub.com&lt;/a&gt;&lt;br&gt;
💬 &lt;strong&gt;Community:&lt;/strong&gt; &lt;a href="https://discord.com/invite/clawd" rel="noopener noreferrer"&gt;Discord&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Feedback, issues, and PRs are all welcome! Would love to hear what you think. 🧡&lt;/p&gt;

</description>
      <category>opensource</category>
      <category>ai</category>
      <category>selfhosted</category>
      <category>javascript</category>
    </item>
  </channel>
</rss>
