AI Bridge
A local multi-agent switchboard for coding agents and CLIs that lets one agent delegate work to another, store job state locally, and feed results back through hooks
- Project type
- Developer Tool
- Role
- Solo Developer
- Date
- Since June 2026
Overview
AI Bridge is an opinionated Hermes-style delegation system for local coding agents and command-line AI tools. It provides a switchboard where one agent can shell out to another agent CLI as a worker, persist the job as local JSON state, and surface the worker result back into the next prompt cycle through hook integrations.
Technology
- Python 3.11
- MCP
- pipx
- JSON
- Shell Hooks
- Codex
- Claude Code
- Cursor Agent
- OpenCode
Key features
- Local multi-agent delegation through ai-delegate and ai-dispatch commands
- First-class support for Codex, Claude Code, Cursor Agent, and OpenCode
- Hook setup command that wires peer inbox hooks across supported agent CLIs
- Local JSON job state for dispatch lifecycle tracking and result handoff
- Auto-routing and route explanation for primary coding agents
- Second-class adapter support for tools such as Goose, Qwen, Gemini, Aider, Amp, Cline, and other shell-callable agents
- MCP peer server entry point for agent coordination
- pipx-friendly installation with multiple command-line entry points
Why it was built
- To coordinate multiple coding agents from one local workflow
- To make delegation between Codex, Claude, Cursor, and OpenCode practical and repeatable
- To experiment with agent orchestration, hook-based inboxes, and local-first automation