opencode-agent-hub

What it is and what it does

opencode-agent-hub is a daemon that enables multiple AI agents running in separate OpenCode sessions to discover and communicate with each other. It works by polling OpenCode's shared SQLite database to find active sessions, facilitating introductions through a coordinator agent, and then routing messages between agents by watching for JSON files written to ~/.agent-hub/messages/ and injecting them into target sessions via async prompts. Agents do not poll for messages—the daemon wakes them up when new communication arrives.

The package is built for developers who want to orchestrate autonomous multi-agent workflows within OpenCode. It includes a daemon service (installable as a systemd or launchd service), a watch utility for monitoring activity, and configurable rate limiting to control LLM API costs. The daemon relies on agent-hub-mcp being configured in OpenCode, and messages are validated with pydantic on receipt. Known limitations include that injected messages are not visible in the OpenCode TUI, and some models may flag orientation messages as prompt injections.

Use it for:

• Orchestrate collaborative problem-solving where multiple specialized agents work together on a single task without manual routing.
• Build autonomous workflows where agents can request information or assistance from peers, reducing external API calls.
• Monitor and debug multi-agent interactions using agent-hub-watch to see message flow and coordination in real time.
• Control LLM API costs in high-frequency agent communication by enabling rate limiting to throttle message throughput.
• Deploy a persistent agent hub as a system service to coordinate long-running agent sessions across multiple terminals.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Coordinates communication between multiple AI agents running in separate OpenCode sessions, enabling them to discover each other and exchange messages through a daemon.

Yes, with conditions. Install if you are actively using OpenCode and want to enable agent-to-agent communication—the low install friction, active maintenance, and zero known vulnerabilities make it safe to try. However, be aware that it is still in Beta, the project is young, and injected messages are not visible in the OpenCode TUI, which may complicate debugging. Verify that agent-hub-mcp is properly configured before starting the daemon, and enable rate limiting if concerned about LLM API costs.

Install

opencode-agent-hub on PyPI

pip install opencode-agent-hub

uv add opencode-agent-hub

poetry add opencode-agent-hub

Installing opencode-agent-hub

Before you install

Low install friction with a pure Python wheel and four lightweight runtime dependencies (prometheus-client, pydantic, requests, watchdog). Active maintenance with a recent release and ongoing commits, though the project is young and still in Beta status.

License in practice

Licensed under AGPL-3.0-only, which requires that any modifications or derivative works be released under the same license and made available to users. This is a strong copyleft license suitable for open-source projects but may restrict commercial use or closed-source modifications.

Quickstart

pip install opencode-agent-hub

# Terminal 1: start the daemon
agent-hub-daemon

# Terminal 2: monitor activity
agent-hub-watch

Requires agent-hub-mcp to be configured in OpenCode's opencode.json before the daemon will start. Python >= 3.11 required.

Verify before relying

• Whether injected agent-to-agent messages consume additional LLM API tokens beyond the agent's own responses.
• Performance characteristics and latency of message injection across many concurrent agents.
• Compatibility with OpenCode versions and whether upstream issue #8564 (injected messages not visible in TUI) has been resolved.
• Whether rate limiting is sufficient for production use cases with high-frequency agent communication.

Package facts

Evidence: opencode_agent_hub-1.6.1-py3-none-any.whl

Keywords: agents, ai, coordination, daemon, mcp, opencode

Tags

More Python Modules packages

Further reading