pac-runner
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@pac-runnerrun the video-production task in examples/video-production.pac.md"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
PAC Runner
PAC Runner 是一个配置化、可发布、可扩展的 Prompt-as-Code 多场景执行器。
它把 QClaw、WorkBuddy、Cursor、Claude Code 或其他智能体的任务,统一转换成标准任务,再交给本地 Codex 执行,并通过 CLI、MCP、任务文件和结果文件形成完整闭环。
核心能力
多场景任务执行:通用、代码、调研、文档、视频制作、交付。
标准任务文件:
.pac.md/.json。标准结果文件:
<任务ID>_结果.md。CLI 入口:
pac-runner。MCP 入口:
pac_run、pac_list、pac_validate。配置化:支持
config.json和环境变量,不依赖个人路径。可扩展:新增模板、场景和工具无需改核心代码。
Related MCP server: Mini Task MCP Server
安装
npm install
npm link或者直接运行:
node bin/pac.mjs list配置
复制配置模板:
cp config.example.json config.json常用环境变量:
export CODEX_EXE="/path/to/codex"
export CODEX_HOME="$HOME/.codex"
export PAC_DEFAULT_CWD="$PWD"
export PAC_DEFAULT_OUTPUT_DIR="$PWD/output"
export PAC_STATE_FILE="$PWD/.pac-runner/state.json"快速使用
查看场景:
pac-runner list直接执行:
pac-runner run --scenario coding --prompt "整理当前目录的脚本说明" --json用标准任务文件:
pac-runner run --task examples/video-production.pac.md --json生成模板:
pac-runner template --scenario video-production --out my-video.pac.md校验任务文件:
pac-runner validate --task examples/video-production.pac.md --jsonMCP 接入
启动 MCP server:
node src/mcp-relay.mjs通用 MCP 配置:
{
"mcpServers": {
"pac-runner": {
"command": "node",
"args": ["/absolute/path/to/pac-runner/src/mcp-relay.mjs"]
}
}
}MCP 工具:
pac_list:列出场景。pac_run:执行标准任务。pac_validate:校验任务文件。
场景
场景 | 用途 |
| 通用分析、整理、执行 |
| 代码、脚本、自动化 |
| 调研、分析、根因排查 |
| Markdown、Word、PPT、Excel、PDF |
| 短视频、分镜、渲染 |
| 任务交付与回传 |
任务文件
---
id: video_task_20260901_1200
scenario: video-production
title: 制作一条15秒短视频
status: ready
priority: normal
owner: pac-runner
cwd: .
outputDir: ./output
inputFiles: []
imagePaths: []
expectedOutputs: ["分镜脚本", "视频文件", "渲染日志"]
timeoutMs: 1200000
---
在这里写具体任务要求。结果文件
执行成功后自动生成:
<outputDir>/<任务ID>_结果.md安全模式
QClaw 技能包调用 Codex 时使用:
--approve-for-me不绕过审批和沙箱,输出目录通过 --add-dir 授权写入。该模式已通过 QClaw 开放平台安全检测。
商业化
商业模式、定价和 Pro/Enterprise 边界见:
测试
npm testThis server cannot be deployed
Maintenance
Related MCP Connectors
- DazbenchOAuthapp.dazbench
Task management your AI agents can actually run. One line becomes a context-ready task over MCP.
Goal and task planning MCP for Codex and AI agents, with evidence-backed completion.
Local-first task manager: create, edit, and complete tasks, projects, and checklists via MCP.
Manage portable AI agent playbooks, Agent Skills, MCP configurations, personas, and memory.
Related MCP Servers
- AlicenseNot gradedqualityDmaintenanceEnables running and managing automated tasks with retry loops and machine-checkable success criteria via MCP tools.25 npmMIT
- FlicenseBqualityCmaintenanceEnables task management via MCP tools for creating, listing, updating, completing, and deleting tasks with JSON file storage.6-
- AlicenseNot gradedqualityCmaintenanceEnables Grok Build to orchestrate the local Codex CLI for code reviews, adversarial reviews, task rescue, session transfer, and background job management through MCP tools.1Apache 2.0
- AlicenseNot gradedqualityAmaintenanceEnables ChatGPT Web or another MCP client to inspect and develop a local workspace, read local Codex task history, and use the complete installed XcodeBuildMCP catalog through a small set of stable tools.MIT