$ loading_
Let Codex delegate coding tasks to worker agents in isolated Git worktrees.
Copy the install command and let the AI configure it · recommended for beginners
No copy-paste install info for "codex-agent-bridge" yet — see the docs or source repo.
Use codex-agent-bridge to start a Claude Code worker in an isolated Git worktree and implement login rate limiting. First use bridge_run to assign the task, then use bridge_respond to follow up on details, and finally use bridge_apply to safely integrate the changes.
A delegated agent workflow is completed, the worker returns results, and approved changes are safely integrated into the main workspace.
Use codex-agent-bridge to create two isolated worktrees so DeepSeek Harness and Claude Code can each implement the same caching module. Compare the results and apply only the better version.
You get independent implementations from different worker agents and can safely integrate the preferred one.
Have Codex use codex-agent-bridge to delegate a refactoring task to a worker agent, but do not merge immediately. Review the changes returned through bridge_run and bridge_respond first, then execute bridge_apply after approval.
Agent output is produced for review first, and changes are safely applied only after approval.
Developers can use Codex as a supervisor and delegate concrete coding tasks to Claude Code or DeepSeek Harness. This separates orchestration from execution and improves collaboration on complex work.
When teams want to avoid agents directly affecting the current workspace, they can run worker agents in isolated Git worktrees. This fits a generate-review-integrate workflow.
When merge risk needs to be controlled, users can integrate worker results through bridge_apply with safe cherry-pick style handling. This is useful when agent-written code should be reviewed before entering the main flow.
It is an MCP server that lets Codex supervise Claude Code or DeepSeek Harness as worker agents. It supports a delegation and integration workflow built around bridge_run, bridge_respond, and bridge_apply.
The description says worker agents run in isolated Git worktrees and their output is integrated through safe cherry-pick handling. That means execution is isolated first, and changes can be applied later by choice.
The provided material does not include installation steps, runtime details, or key requirements. For more information, see the source repository.
Delegate implementation tasks from AI coding agents to the Codex CLI.
Delegate long-running Codex agent tasks to Antigravity CLI with observable execution.
Coordinate Codex and Claude Code through a shared task board and inbox.
Delegate tasks between Claude and ChatGPT/Codex agents with queues and security controls.
Bridge Claude with local Codex skills and plugins for task execution.
Connect ChatGPT web to local Codex for approved file access and tasks.