claude-codex-orchestrator
A local MCP server that manages Claude-to-Codex task orchestration on Windows, providing tools for creating, dispatching, and monitoring tasks with safety constraints.
README
Claude Codex Orchestrator MCP
A local MCP server that wraps a Windows Claude -> Codex worker orchestration workflow.
The MCP server is intentionally thin. It exposes a stable tool surface and delegates execution to the existing PowerShell dispatcher scripts under tools/.
What It Provides
create_task: create one queued markdown task under.orchestrator/tasks/.dispatch_task: dispatch one task throughtools/invoke-orchestrated-task.ps1.read_status: read compactstatus.jsonoutput and summarize the result.run_task_loop: run bounded retry/decision handling over queued tasks.archive_artifacts: archive historical artifacts, dry-run by default.
Requirements
- Windows PowerShell 5.x or compatible
powershell.exe. - Node.js 20+.
- Codex CLI available on
PATHforappandcodelanes. - Codex Desktop/app-server support for the
applane.
Quick Start
npm run mcp:orchestrator
The server speaks MCP over stdio. It is meant to be launched by an MCP client, not used as an interactive terminal program.
MCP Client Config Example
Use absolute paths for your clone location and the workspace you want the orchestrator to manage:
{
"mcpServers": {
"claude-codex-orchestrator": {
"command": "C:/Program Files/nodejs/node.exe",
"args": [
"D:/path/to/claude-codex-orchestrator-mcp/tools/mcp-orchestrator-server.js"
],
"env": {
"ORCHESTRATOR_WORKSPACE_ROOT": "D:/path/to/your/workspace"
}
}
}
}
Safety Model
- The server validates file paths and keeps them inside the configured workspace.
- Set
ORCHESTRATOR_WORKSPACE_ROOTwhen the MCP server is installed outside the workspace it should manage. - Automation defaults to dry-run unless explicitly invoked with real automation flags in the underlying dispatcher path.
- Artifact archival defaults to dry-run; moves require
apply: true. - The MCP server does not edit Codex global config.
- Existing PowerShell scripts remain the source of truth for orchestration behavior.
Repository Layout
.
|-- AGENTS.md
|-- PROJECT-CONTEXT-MCP-SKILL.md
|-- package.json
|-- docs/
| `-- ORCHESTRATOR-USAGE.md
|-- tools/
| |-- mcp-orchestrator-server.js
| |-- invoke-orchestrated-task.ps1
| |-- invoke-codex-app-worker.ps1
| |-- codex-app-worker.js
| |-- invoke-codex-worker.ps1
| |-- run-controlled-automation.ps1
| |-- orchestrator-tasking.ps1
| |-- invoke-orchestrator-task-loop.ps1
| `-- archive-orchestrator-artifacts.ps1
`-- .claude/
|-- commands/orchestrate.md
`-- skills/claude-codex-orchestrator/SKILL.md
Verification
Syntax-check the MCP server:
node --check tools/mcp-orchestrator-server.js
Run non-Codex PowerShell checks:
powershell -NoProfile -ExecutionPolicy Bypass -File tools/test-orchestrator-tasking.ps1
powershell -NoProfile -ExecutionPolicy Bypass -File tools/test-controlled-automation.ps1
powershell -NoProfile -ExecutionPolicy Bypass -File tools/test-archive-orchestrator-artifacts.ps1
Full app/code worker tests require a working Codex CLI/app-server setup and may take longer.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。