claude-chatgpt-agent-bridge
An MCP task broker that enables Claude and Codex/ChatGPT agents to delegate tasks to each other, with persistent queues, worker adapters, and security controls.
README
Claude + ChatGPT Agent Bridge
Agent Bridge is a working MCP task broker that lets Claude and Codex/ChatGPT agents assign bounded work to one another, continue independently, and retrieve the peer result later.
The linked Cognito Forms video demonstrates a useful shared-tool pattern: both AI products connect to one external MCP-backed system. Agent Bridge extends that pattern with a task queue and worker adapters, because sharing a connector alone does not create agent-to-agent delegation.
Documentation
New here? Start with the User Guide. See also the Development workflow, Setup Tutorial, and Validation and Compatibility.
Current Status
- Working MCP server over local stdio and remote-capable Streamable HTTP
- Persistent asynchronous task queue
- Claude Code and Codex CLI worker adapters
- Read-only and workspace-write modes
- Workspace allowlist instead of arbitrary paths
- Cancellation, restart recovery, nesting limits, and circular-delegation protection
- Static bearer authentication for local/private setups
- OAuth/JWT protected-resource support for hosted connectors
- Shared Claude/Codex delegation skill and plugin scaffolds
- Automated protocol and queue tests
- Real authenticated Codex-to-Claude and Claude-to-Codex Windows handoffs completed
See Validation and Compatibility for what is proven, what depends on account access, and the product-by-product support matrix. Follow Setup Tutorial to install it.
Quick Local Start
On Windows, macOS, or Linux, the repeatable setup command checks the prerequisites, builds and tests the bridge, creates the local config if needed, and registers the MCP endpoint with both CLIs:
node scripts/setup-local.mjs
To make a particular project the default allowed workspace during setup:
node .\scripts\setup-local.mjs --workspace-name my-project --workspace "C:\path\to\project"
Existing config and MCP registrations are preserved. To perform the same steps manually:
npm install
npm run build
cp bridge.config.example.json bridge.config.json
npm run doctor
Edit bridge.config.json so its named workspaces point to the project folders the agents may access. Start the single local broker before opening either agent:
npm run start:http
Windows can run the broker detached with npm run start:windows and stop it with npm run stop:windows. The setup tutorial also shows the optional sign-in startup entry.
Connect both Claude Code and Codex to http://127.0.0.1:8787/mcp using the examples in configs/. One shared HTTP process is the recommended two-agent setup. Stdio is included for single-client or diagnostic use.
The MCP tools are:
agent_bridge_capabilitiesdelegate_taskget_tasklist_taskscancel_task
Security Model
The bridge never asks for or copies ChatGPT or Claude session tokens. Local workers reuse the authenticated claude and codex CLIs on the same machine. HTTP deployment requires a bearer token or OAuth/JWT configuration when it binds outside loopback. Provider API/CLI credentials remain on the worker host and are not returned through MCP.
See Security before exposing the HTTP endpoint.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
E2B
Using MCP to run code via e2b.