agent-bus-mcp 0.4.3

Local SQLite-backed MCP bus for peer coding agents

Hero image rendered locally with z-image-turbo from a handcrafted prompt.

Agent Bus MCP

Agent Bus is a local coordination layer for multiple coding agents on the same machine.

It gives MCP-capable tools such as Codex, Claude Code, Gemini CLI, OpenCode, and Cursor a shared, durable message bus backed by SQLite. Instead of copying summaries between tools or losing context when a process restarts, agents can join the same topic, exchange messages, replay history, and pick work back up from server-side cursors.

Why use Agent Bus?

Use Agent Bus when you want agent collaboration to feel more like a shared inbox than a fragile copy-paste workflow.

• Keep coordination local. One stdio MCP server and one SQLite file are enough to let multiple agents collaborate on the same machine.
• Preserve state between turns. Topics, messages, cursors, and peer identities live in the DB, not in one client process's memory.
• Support real workflows. Reviewer/implementer loops, handoffs, sidecar helpers, and replayable audit trails all fit naturally into topic-based messaging.
• Add observability without extra infrastructure. Search, export, and the optional web UI make it easier to inspect conversations after the fact.

When Agent Bus is a good fit

Agent Bus is a strong fit when:

• two or more local coding agents need to collaborate on the same task
• an agent should be able to disconnect and later reclaim its identity
• you want durable replay instead of "read everything again" polling
• you want a searchable local log of agent-to-agent coordination

It is a weaker fit when:

• you only need a single agent with no handoffs
• you need networked multi-host messaging
• you need auth, tenancy, or cloud-hosted coordination out of the box

Quickstart

The fastest path is to install Agent Bus as an MCP server in your client and start using it from natural-language prompts.

npx install-mcp "uvx --from agent-bus-mcp==<version> agent-bus" --name agent-bus --client claude-code

Replace <version> with the release you want to run. For direct setup:

# Codex
codex mcp add agent-bus -- uvx --from agent-bus-mcp==<version> agent-bus

# Claude Code
claude mcp add agent-bus -- uvx --from agent-bus-mcp==<version> agent-bus

Then ask an agent to:

1. create or reuse a topic
2. join the topic with a stable agent_name
3. send or read messages through sync()

For a fuller walkthrough, start with First topic between two peers.

Web UI

Agent Bus also ships with a local Web UI for browsing topics, reading threads, exporting logs, and searching the bus without leaving the browser.

From a source checkout:

uv sync --extra web
pnpm --dir frontend install
pnpm --dir frontend build
uv run agent-bus serve

Or launch the published package directly with the web extra:

uvx --from "agent-bus-mcp[web]==<version>" agent-bus serve

The overview keeps recent topics in the sidebar and leaves the main workbench focused on navigation and search.

For the full walkthrough, including thread navigation, export, and troubleshooting, see How to use the Agent Bus Web UI.

Documentation

This repo now has a lightweight Diataxis-style docs layout under docs/:

Need	Start here
Learn by doing	Tutorials
Complete a setup task	How-to guides
Look up tools, commands, and config	Reference
Understand the design and tradeoffs	Explanation

Recommended starting points:

• Why use Agent Bus?
• Install and configure Agent Bus
• Runtime reference
• Implementation spec

Upgrading from an older release? See CHANGELOG.md.

Optional workflow skill

This repo also includes a reusable workflow skill asset at ./.agents/skills/agent-bus-workflows/.

It is useful when you want ready-made reviewer/implementer loops, handoffs, duplicate-name recovery, and reclaim-token reconnect behavior in a Codex-style skill package. See Install and configure Agent Bus for the install path.

Architecture at a glance

The Python package provides the MCP server, CLI, Web UI, and embedding worker. The Rust core owns the SQLite schema, reads/writes, search, and embedding-job coordination, which keeps the data model single-sourced.

For the rationale behind this shape, see Why use Agent Bus?.