GUIDE

Passfair, in five minutes.

A local app where three AI agents plan, produce and review a task together, correcting each other automatically. You bring your own API keys; nothing is uploaded anywhere except the model calls you authorise.

1. Install

1. Install Node.js 24 or newer (free, one installer).
2. Unzip the Passfair download anywhere, e.g. your Documents folder.
3. Start it: Windows → double-click start-windows.cmd; macOS → right-click start-macos.command → Open → Open (first time only); Linux → ./start-linux.sh.
4. Security prompts on first launch are expected — these are plain scripts, not a signed app. macOS “cannot be opened because it is from an unidentified developer”: System Settings → Privacy & Security → scroll down → Open Anyway. Windows SmartScreen “Windows protected your PC”: More info → Run anyway.
5. Your browser opens http://127.0.0.1:3900. Keep the black window open while you use Passfair; close it to stop.

Apart from Node.js there is no installer, no admin rights are needed and no npm packages are downloaded. Everything lives in the folder you unzipped, including the data/ folder with your jobs and encrypted keys.

2. The screen, in one glance

One row of buttons at the top — Claude · GPT · Gemini · Muse — and one conversation below. Connect an agent by clicking its button and pasting a key. Type what you need in the box at the bottom. The agents talk to each other in the conversation; you can cut in at any time by typing while they work, and the team stops and follows your new instruction. Team (top right) shows who does what; presets are set automatically from the agents you connected.

2b. Watch the demo

With no agent connected, press Watch a 20-second demo (or just type a brief and Send). Synthetic agents walk through the loop — plan, draft, a review that sends it back, a corrected draft, approval. No key, no cost. It shows you exactly what a live run looks like.

3. Add API keys

Click the agent's button — Claude, GPT or Gemini — paste the key, keep or change the model name, press Connect. Passfair makes one tiny test call; the button turns green. Roles are assigned automatically (two agents connected → the Make & ship preset: Claude plans and drafts, GPT polishes, Claude reviews).

• Anthropic (Claude) — console.anthropic.com → API keys. Default model claude-sonnet-4-5.
• OpenAI (GPT) — platform.openai.com → API keys. Default model gpt-4.1.
• Google (Gemini) — aistudio.google.com → Get API key. Default model gemini-2.5-flash.

Model names change over time; if a save fails with “check the model name”, look up the current name in that provider's documentation. Costs are billed by the providers to your own accounts; a typical 3-round job uses 5 to 7 model calls.

4. Run a live job

1. Write the brief: what to produce, for whom, in what shape. Paste any facts into Material — agents cannot browse, so this is their only source of truth.
2. Open Team only if you want to change who does what. Tip: keep the Reviewer on a different vendor than the Worker.
3. Tick Live under the message box and press Send.
4. Watch the agents talk. Each review lists concrete issues; each revised draft answers them. Type at any moment to cut in — the team stops and follows you. When the Reviewer passes it, the Final deliverable appears.
5. Read it critically, then Accept, Copy, Export, or hand it to Muse. To change something, just Reply in the same conversation.

5. Use it from Claude, Cursor or any MCP client

Passfair is also an MCP server, so your chat assistant can hand it a brief and fetch the result — no copy-paste. Add this to your client's MCP configuration (use the absolute path to studio-connector.mjs):

{
  "mcpServers": {
    "agent-studio": {
      "command": "node",
      "args": ["/ABSOLUTE/PATH/agent-studio/studio-connector.mjs"]
    }
  }
}

Passfair must be running. Tools: studio_job_create, studio_job_get, studio_job_list, studio_job_cancel. Example prompt to your assistant: “Create a Passfair live job to draft the Q1 newsletter with these notes, then fetch the final deliverable when it is ready.”

5b. Ship it with Muse (the distributor)

Passfair makes; Muse publishes. Pair them once and the carrying stops being your job.

1. Pair: click the Muse button → Pair with Muse. Passfair shows a one-time code (e.g. K7P-4QX). In Muse, connect “Passfair” and enter it. The status turns to paired.
2. Order from Muse: tell Muse “ask Passfair to write …”. Passfair picks the brief up within a minute, runs the Make & ship team (Claude logic → GPT polish → Claude review) and, when the Reviewer approves, hands the result back automatically (setting Auto hand-off).
3. Muse publishes. If publishing or deploying fails, Muse reports the problem; Passfair re-opens the job, the Worker fixes it, the Reviewer re-checks, and version 2 goes back — you see every step on the timeline.
4. Without the connector (while Meta reviews it): press Copy for Muse on a finished job, paste into Muse, and if Muse reports an error paste it into Publishing hit a problem? — same fix loop, carried by hand.

What travels: only the brief Muse places and the finished deliverable you hand off, through the Agents Fair relay, encrypted, kept at most 7 days, deleted on unpair. API keys and drafts in progress never leave this computer. Muse cannot approve anything in Passfair; you (or the auto hand-off rule you chose) decide what ships.

What Passfair does not do

• It does not browse the web, read files, send email, pay for anything or publish. Agents only write and critique text.
• It does not control ChatGPT, Claude or other apps on your computer, and does not read their history.
• It does not guarantee correctness. A reviewer catches many errors, not all. Verify facts before using the result.
• It stores keys encrypted on this computer only; the operator of your computer can still read them. Do not run it on a shared machine.

中文说明

这是什么：一个在你自己电脑上运行的小程序。你给一个任务，三个 AI 角色自动协作——规划者定标准，执行者写出成果，审核者挑错并退回修改，循环到通过为止。你只看最终版，不用在几个 AI 之间复制粘贴。

安装（Mac / Windows 应用版）：解压后双击 Passfair，浏览器自动打开，不用装别的。退出点页面右上角 Quit。首次打开若有系统提示：Mac 右键 → 打开；Windows 更多信息 → 仍要运行。

安装（脚本版）：① 安装 Node.js 24 以上（nodejs.org，免费）；② 解压下载包；③ Windows 双击 start-windows.cmd（若弹出 SmartScreen "Windows 已保护你的电脑"，点"更多信息 → 仍要运行"）；macOS 右键 start-macos.command → 打开 → 打开（若提示"无法打开"，到 系统设置 → 隐私与安全性 → 仍要打开）；④ 浏览器自动打开 http://127.0.0.1:3900。黑色窗口保持打开，关掉即停止。

先看演示：点 Watch a demo，不需要任何 Key，看完整流程。

接入 AI：顶部一排按钮 Claude · GPT · Gemini · Muse，点一个、粘贴 Key、点 Connect，按钮变绿即接入；接两家后自动按 "Make & ship" 分工（Claude 规划起草、GPT 润色、Claude 审核）。

正式运行：在底部对话框里写任务，勾 Live，发送。几个 AI 在同一个对话框里来回：起草、审核退回、修改、通过。任何时候你都可以直接打字插话——团队会停下来按你的新指示继续。最终成果出现后可复制、导出、交给 Muse，或者直接回复要求修改。右上角 Team 可以看或改分工，一般不用动。

交给 Muse 发布：点 Muse 按钮 → Pair with Muse，得到一次性配对码，在 Muse 里连接 "Passfair" 输入即可。之后在 Muse 里说"让 Passfair 写……"，Passfair 自动领单、按"Make & ship"预设（Claude 负责逻辑 → GPT 润色 → Claude 审核）完成、审核通过后自动交回 Muse 发布；发布出错 Muse 会把问题回传，Passfair 自动重开任务修好再交第二版，全程在时间线上可见。Meta 审核 connector 期间，用成稿区的 Copy for Muse 粘到 Muse，出错把错误贴回 "Publishing hit a problem?"，修复循环一样。

接入 Claude / Cursor 等：Passfair 同时是一个 MCP 服务，把上面第 5 节的配置加进你的客户端，就能对助手说"把这个任务交给 Passfair 做，做完把结果拿回来"。

不做什么：不上网、不读文件、不发邮件、不付款、不发布；不控制你电脑上的 ChatGPT 或 Claude 应用；不保证结果绝对正确——审核能抓很多错，但不是全部，事实请自行核对。费用由各厂商按你自己的账号计费，一次 3 轮任务通常 5–7 次模型调用。