codex-agent-bridge
Click on "Install 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., "@codex-agent-bridgebridge_run claude to implement password reset in auth module"
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.
Agent Bridge
Codex supervises Claude Code and DeepSeek Harness. Bridge itself does not call LLM.
Let Codex be the only supervisor. Claude Code or DeepSeek Harness does the implementation. You review, verify, and apply in Codex.
One Command
npx -y codex-agent-bridgeThen start a new Codex session. You will see the bridge_* tools.
This command will: install the runtime to ~/.agent-bridge, write Codex Skill and MCP (using local node.exe, not npx.cmd), and run doctor once.
To update, run the same command again. To uninstall:
npx -y codex-agent-bridge uninstall
npx -y codex-agent-bridge uninstall --purge--purge will delete ~/.agent-bridge.
Prerequisites
Required | Description |
Node.js 22+, Git | To run Bridge |
Codex | Only Supervisor |
Claude Code or DeepSeek Harness | At least one that can log in and work independently |
We don't handle
Won't install Claude Code / DeepSeek Harness for you
Won't change their model / provider / reasoning effort
Won't call any LLM
Whichever Worker is available, that one will be dispatched. When none are available, doctor will indicate.
Related MCP server: cc-agent
What it does
Codex ──bridge_run──► Claude Code / DeepSeek Harness
▲ │
│ ReviewPacket │ 在独立 worktree 里改代码
│ verification │ 不得 commit / push / merge
└──── bridge_apply ◄──────┘Closed loop:
bridge_run
→ WAITING_FOR_INPUT 时 bridge_respond
→ AWAITING_REVIEW + ReviewPacket
→ bridge_approve 任务分支 checkpoint,不是完成发布
→ bridge_apply cherry-pick 到当前分支,不是 mergeTurn end ≠ task completion. Diff only recognizes baseCommit vs worktree.
bridge_run only selects worker: claude or deepseek (debugging can use replay / fake). Do not pass model / provider / effort.
If DeepSeek Harness is not in a common path, set AGENT_BRIDGE_DEEPSEEK_ROOT.
Usage Notes
The target must be an independent git repository. Each
bridge_*call carries the sameproject. The task UUID field name istask.Only one Writer per project at a time. A second writer returns
CORE_LOCK_HELD.After modifying MCP, you must start a new Codex session; old sessions won't see the new tools.
A corrupted
tasks.jsonwill result inTASK_STORE_CORRUPTED, not treated as an empty store.
Publishing a New Version
Local npm publish always requires 2FA; this is an npm rule and cannot be made passwordless like git push.
The correct approach is: change version → tag → push, and GitHub Actions uses Trusted Publishing to publish to npm, no password needed.
One-time setup (only needed once): Open codex-agent-bridge npm settings → Trusted Publisher → GitHub Actions, fill in:
Field | Value |
Organization or user |
|
Repository |
|
Workflow filename |
|
Allowed actions |
|
After that, each time:
# 先改 package.json 里的 version,提交
git tag v1.2.0
git push origin main --tagsDeveloping from Source
npm install
npx tsx src/cli.ts setup --dev
npm test--dev makes MCP point to the current repository instead of ~/.agent-bridge.
Documentation
Product Roadmap | |
How to integrate with Codex | |
Why npm instead of EXE | |
Why not HTTP/GUI |
The repository directory may still be called Agent Relay. The product name is Agent Bridge. The npm package name is codex-agent-bridge.
This server cannot be installed
Maintenance
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Related MCP Servers
- AlicenseAqualityCmaintenanceA local MCP server that lets Claude delegate scoped work to Codex with structured results and guardrails, supporting planning, code review, build, reverse engineering, and long-running background tasks.11MIT
- AlicenseAqualityBmaintenanceMCP server that spawns autonomous Claude Code agents in GitHub repos, enabling task delegation with persistent state, multi-step workflows, and job monitoring.471632Apache 2.0
- AlicenseNot gradedqualityBmaintenanceMCP server enabling Claude to consult Codex (GPT-5.x) mid-task for second opinions, plan/diff review, brainstorming, and codebase exploration via structured debates and permission-controlled interactions.2MIT
- AlicenseNot gradedqualityCmaintenanceAn MCP server that lets ChatGPT or any MCP client securely delegate coding tasks to a local Claude Code instance, with git checkpointing, approval gates, and structured results. Supports code review, test running, and rollback via simple tool calls.16MIT
Related MCP Connectors
Hosted MCP server connecting claude.ai, ChatGPT and other AI apps to your own computer
An MCP server that gives your AI access to the source code and docs of all public github repos
A paid remote MCP for OpenAI Codex agent coordination MCP, built to return verdicts, receipts, usage
Latest Blog Posts
- Who's Calling? MCP Hosts Are an Identity Blind Spot (And the Spec Knows It)By Om-Shree-0709 on .mcpAgent IdentityOAuth 2.1
- Your AI Chatbot Just Exposed Your CEO's Salary to an InternBy Om-Shree-0709 on .Agent IdentityMCP SecurityOAuth Delegation
- Why MCP Servers Need Execution Sandboxing (And Why Your Current Stack Isn't Enough)By Om-Shree-0709 on .Agentic AiPrompt InjectionWebAssembly
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/JaylanJerry/codex-agent-bridge'
If you have feedback or need assistance with the MCP directory API, please join our Discord server