AgentBridge

Local bridge for bidirectional communication between Claude Code and Codex inside the same working session.

AgentBridge uses a two-process architecture:

bridge.ts is the foreground MCP client started by Claude Code via the AgentBridge plugin
daemon.ts is a persistent local background process that owns the Codex app-server proxy and bridge state

When Claude Code closes, the foreground MCP process exits while the background daemon and Codex proxy keep running. When Claude Code starts again, it reconnects automatically with exponential backoff.

What this project is / is not

This project is:

A local developer tool for connecting Claude Code and Codex in one workflow
A bridge that forwards messages between an MCP channel and the Codex app-server protocol
An experimental setup for human-in-the-loop collaboration between multiple agents

This project is not:

A hosted service or multi-tenant system
A generic orchestration framework for arbitrary agent backends
A hardened security boundary between tools you do not trust

Architecture

┌──────────────┐     MCP stdio / plugin     ┌────────────────────┐
│ Claude Code  │ ──────────────────────────▶ │ bridge.ts          │
│ Session      │ ◀──────────────────────────  │ foreground client  │
└──────────────┘                             └─────────┬──────────┘
                                                       │
                                                       │ control WS (:4502)
                                                       ▼
                                             ┌────────────────────┐
                                             │ daemon.ts          │
                                             │ bridge daemon      │
                                             └─────────┬──────────┘
                                                       │
                                     ws://127.0.0.1:4501 proxy
                                                       │
                                                       ▼
                                             ┌────────────────────┐
                                             │ Codex app-server   │
                                             └────────────────────┘

Data flow

Direction	Path
Codex -> Claude	`daemon.ts` captures `agentMessage` -> control WS -> `bridge.ts` -> `notifications/claude/channel`
Claude -> Codex	Claude calls the `reply` tool -> `bridge.ts` -> control WS -> `daemon.ts` -> `turn/start` injects into the Codex thread

Loop prevention

Each message carries a source field ("claude" or "codex"). The bridge never forwards a message back to its origin.

Prerequisites

Dependency	Version	Install
Bun	v1.0+	`curl -fsSL https://bun.sh/install \| bash`
Claude Code	v2.1.80+	`npm install -g @anthropic-ai/claude-code`
Codex CLI	latest	`npm install -g @openai/codex`

Note: Bun is required as the runtime for the AgentBridge daemon and plugin server. Node.js alone is not sufficient.

Quick Start

Install via Plugin Marketplace (recommended)

Install AgentBridge directly from Claude Code using the plugin marketplace:

# 1. In Claude Code, add the AgentBridge marketplace
/plugin marketplace add raysonmeng/agent-bridge

# 2. Install the plugin
/plugin install agentbridge@agentbridge

# 3. Reload plugins to activate
/reload-plugins

Then install the CLI tool:

# 4. Install the CLI globally
npm install -g @raysonmeng/agentbridge

# 5. Generate project config (optional)
abg init

# 6. Start Claude Code with AgentBridge channel enabled
abg claude

# 7. Start Codex TUI connected to the bridge (in another terminal)
abg codex

Tip: abg is a short alias for agentbridge. Both commands are identical — use whichever you prefer.

That's it. The daemon starts automatically when needed and reconnects if restarted.

Updating the plugin

When a new version is released, update from Claude Code:

/plugin marketplace update agentbridge
/reload-plugins

Or enable auto-update: run /plugin → Marketplaces tab → select agentbridge → Enable auto-update.

Install for local development

If you want to modify AgentBridge source code, use the local development setup instead:

# 1. Clone and install dependencies
git clone https://github.com/raysonmeng/agent-bridge.git
cd agent-bridge
bun install
bun link

# 2. Set up local plugin + project config
agentbridge dev     # Register local marketplace + install plugin
agentbridge init    # Check dependencies, generate .agentbridge/config.json

# 3. Start Claude Code with AgentBridge plugin loaded
agentbridge claude

# 4. Start Codex TUI connected to the bridge (in another terminal)
agentbridge codex

Note: agentbridge claude injects --dangerously-load-development-channels plugin:agentbridge@agentbridge automatically. This loads a local development channel into Claude Code (currently a Research Preview workflow). Only enable channels and MCP servers you trust.

Updating after code changes

After modifying AgentBridge source code, re-run agentbridge dev to sync changes to the plugin cache, then restart Claude Code or run /reload-plugins in an active session.

CLI Reference

All commands work with both agentbridge and the short alias abg.

Command	Description
`abg init`	Install plugin, check dependencies (bun/claude/codex), generate `.agentbridge/config.json`
`abg claude [args...]`	Start Claude Code with push channel enabled. Clears any killed sentinel from a previous `kill`. Pass-through args are forwarded to `claude`
`abg codex [args...]`	Start Codex TUI connected to AgentBridge daemon. Manages TUI process lifecycle (pid tracking, cleanup). Pass-through args forwarded to `codex`
`abg kill`	Gracefully stop both daemon and managed Codex TUI, clean up state files, write killed sentinel
`abg dev`	(Dev only) Register local marketplace + force-sync plugin to cache
`abg --help`	Show help
`abg --version`	Show version

Owned flags

Some flags are automatically injected and cannot be manually specified:

agentbridge claude owns: --channels, --dangerously-load-development-channels
agentbridge codex owns: --remote, --enable tui_app_server

Passing these flags manually will result in a hard error with guidance to use the native command directly.

Note on flag positioning for agentbridge codex: For the bare TUI form (agentbridge codex …), bridge flags are injected at the front. For TUI subcommands that carry per-subcommand args (resume, fork), they are injected after the subcommand name (so clap parses them as options of the actually-invoked command, not the root). Non-TUI subcommands like exec, mcp, plugin, remote-control, update etc. are passed through unchanged — no bridge flags injected. See src/cli/codex.ts buildCodexArgs for the full positioning logic.

Project Config

Running agentbridge init creates a .agentbridge/ directory in your project root:

File	Purpose
`config.json`	Machine-readable project config (Codex ports, turn coordination, idle shutdown)

The config is loaded by the CLI and daemon at startup. Re-running init is idempotent and will not overwrite existing files.

File Structure

agent_bridge/
├── .github/
│   ├── ISSUE_TEMPLATE/           # Bug report and feature request templates
│   ├── pull_request_template.md
│   └── workflows/ci.yml          # GitHub Actions CI
├── assets/                        # Static assets (images, etc.)
├── docs/
│   ├── phase3-spec.md            # Phase 3 design spec (CLI + Plugin)
│   ├── v1-roadmap.md             # v1 feature roadmap
│   └── v2-architecture.md        # v2 multi-agent architecture design
├── plugins/agentbridge/           # Claude Code plugin bundle
│   ├── .claude-plugin/plugin.json
│   ├── commands/init.md
│   ├── hooks/hooks.json
│   ├── scripts/health-check.sh
│   └── server/                    # Bundled bridge-server.js + daemon.js
├── src/
│   ├── bridge.ts                  # Claude foreground MCP client (plugin entry point)
│   ├── daemon.ts                  # Persistent background daemon
│   ├── daemon-client.ts           # WebSocket client for daemon control port
│   ├── daemon-lifecycle.ts        # Shared daemon lifecycle (ensureRunning, kill, startup lock)
│   ├── control-protocol.ts        # Foreground/background control protocol types
│   ├── claude-adapter.ts          # MCP server adapter for Claude Code channels
│   ├── codex-adapter.ts           # Codex app-server WebSocket proxy and message interception
│   ├── config-service.ts          # Project config (.agentbridge/) read/write
│   ├── state-dir.ts               # Platform-aware state directory resolver
│   ├── message-filter.ts          # Smart message filtering (markers, summary buffer)
│   ├── types.ts                   # Shared types
│   ├── cli.ts                     # CLI entry point and command router
│   └── cli/
│       ├── init.ts                # agentbridge init
│       ├── claude.ts              # agentbridge claude
│       ├── codex.ts               # agentbridge codex
│       ├── kill.ts                # agentbridge kill
│       └── dev.ts                 # agentbridge dev
├── CLAUDE.md                      # Project rules for AI agents
├── CODE_OF_CONDUCT.md
├── CONTRIBUTING.md
├── LICENSE
├── README.md
├── README.zh-CN.md
├── SECURITY.md
├── package.json
└── tsconfig.json

Configuration

Environment Variables

Variable	Default	Description
`CODEX_WS_PORT`	`4500`	Codex app-server WebSocket port
`CODEX_PROXY_PORT`	`4501`	Bridge proxy port for the Codex TUI
`AGENTBRIDGE_CONTROL_PORT`	`4502`	Control port between bridge.ts and daemon.ts
`AGENTBRIDGE_LIVENESS_PROBE_TIMEOUT_MS`	`3000`	Maximum wait for incumbent Claude pong before evicting on contention (issue #68)
`AGENTBRIDGE_STATE_DIR`	Platform default	State directory for pid, status, logs (macOS: `~/Library/Application Support/agentbridge/`, Linux: `$XDG_STATE_HOME/agentbridge/`)
`AGENTBRIDGE_MODE`	`push`	Message delivery mode (`push` for channels, `pull` for API key mode)
`AGENTBRIDGE_DAEMON_ENTRY`	`./daemon.ts`	Override daemon entry point (used by plugin bundles)

State Directory

The daemon stores runtime state in a platform-aware directory:

Platform	Default Path
macOS	`~/Library/Application Support/agentbridge/`
Linux	`$XDG_STATE_HOME/agentbridge/` (fallback: `~/.local/state/agentbridge/`)

Contents: daemon.pid, status.json, agentbridge.log, killed (sentinel), startup.lock

Disabled Bridge States

The bridge can enter several dormant states when it cannot accept new MCP replies. Each state surfaces to the agent as an error message (and, for the transient ones, an in-band push notification):

State	Cause	Recovery
`killed`	`agentbridge kill` was run, sentinel file present.	Restart Claude Code (`agentbridge claude`), switch to a new conversation, or run `/resume`.
`rejected`	Daemon rejected the connection: another Claude session is already attached.	Close the other session, or run `agentbridge kill` to reset, then `agentbridge claude` again.
`evicted`	A newer session evicted this one after the incumbent failed a liveness probe (issue #68).	Close this session and start a fresh one with `agentbridge claude`.
`probe_in_progress`	A liveness probe is currently checking the incumbent — contention window. Transient (auto-recovers within `DISABLED_RECOVERY_INTERVAL_MS` × cap, ~30 s).	None needed; the recovery poller reconnects automatically when the slot clears.
`auto_recovery_exhausted`	The auto-recovery poller for `probe_in_progress` ran its full retry budget (6 attempts, ~30 s) without succeeding. Terminal.	Retry manually with `agentbridge claude`.

Current Limitations

Only forwards agentMessage items, not intermediate commandExecution, fileChange, or similar events
Single Codex thread, no multi-session support yet
Single Claude foreground connection; a new Claude session replaces the previous one
Fixed ports mean only one AgentBridge instance per machine (multi-project support planned for post-v1)

Codex git restrictions

Codex runs in a sandboxed environment that blocks all writes to the .git directory. This means Codex cannot run git commit, git push, git pull, git checkout -b, git merge, or any other command that modifies git metadata. Attempting these commands will cause the Codex session to hang indefinitely.

Recommendation: Let Claude Code handle all git operations (branching, committing, pushing, creating PRs). Codex should focus on code changes and report completed work via agentMessage, then Claude Code takes care of the git workflow.

Roadmap

v1.x (current): Improve the single-bridge experience without architectural refactoring -- less noise, better turn discipline, and clearer collaboration modes. See docs/v1-roadmap.md.
v2 (planned): Introduce the multi-agent foundation -- room-scoped collaboration, stable identity, a formal control protocol, and stronger recovery semantics. See docs/v2-architecture.md.
v3+ (longer term): Explore smarter collaboration, richer policies, and more advanced orchestration across runtimes.

How This Project Was Built

This project was built collaboratively by Claude Code (Anthropic) and Codex (OpenAI), communicating through AgentBridge itself -- the very tool they were building together. A human developer coordinated the effort, assigning tasks, reviewing progress, and directing the two agents to work in parallel and review each other's output.

In other words, AgentBridge is its own proof of concept: two AI agents from different providers, connected in real time, shipping code side by side.

Contact

This is my first open-source project! I'd love to connect with anyone interested in multi-agent collaboration, AI tooling, or just building cool things together. Feel free to reach out:

Twitter/X: @raysonmeng
Xiaohongshu: Profile
WeChat: Scan the QR code below to add me

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

AgentBridge

What this project is / is not

Architecture

Data flow

Loop prevention

Prerequisites

Quick Start

Install via Plugin Marketplace (recommended)

Updating the plugin

Install for local development

Updating after code changes

CLI Reference

Owned flags

Project Config

File Structure

Configuration

Environment Variables

State Directory

Disabled Bridge States

Current Limitations

Codex git restrictions

Roadmap

How This Project Was Built

Contact

About

Uh oh!

Releases 6

Packages

Uh oh!

Contributors

Uh oh!

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 63 Commits
.claude-plugin		.claude-plugin
.github		.github
assets		assets
docs		docs
plugins/agentbridge		plugins/agentbridge
scripts		scripts
src		src
.gitignore		.gitignore
.mcp.json.example		.mcp.json.example
CLAUDE.md		CLAUDE.md
CODE_OF_CONDUCT.md		CODE_OF_CONDUCT.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
README.zh-CN.md		README.zh-CN.md
SECURITY.md		SECURITY.md
bun.lock		bun.lock
package.json		package.json
tsconfig.json		tsconfig.json

Folders and files

Latest commit

History

Repository files navigation

AgentBridge

What this project is / is not

Architecture

Data flow

Loop prevention

Prerequisites

Quick Start

Install via Plugin Marketplace (recommended)

Updating the plugin

Install for local development

Updating after code changes

CLI Reference

Owned flags

Project Config

File Structure

Configuration

Environment Variables

State Directory

Disabled Bridge States

Current Limitations

Codex git restrictions

Roadmap

How This Project Was Built

Contact

About

Topics

Resources

License

Code of conduct

Contributing

Security policy

Uh oh!

Stars

Watchers

Forks

Releases 6

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages