![\"agent-bus](\"docs/assets/banner.png\")
\n\n \n
\n ![\"npm](\"https://img.shields.io/npm/v/@agent-bus-connect/cli.svg?label=npm\")
\n ![\"license\"](\"https://img.shields.io/npm/l/@agent-bus-connect/cli.svg\")
\n ![\"downloads\"](\"https://img.shields.io/npm/dm/@agent-bus-connect/cli.svg\")
\n ![\"Agent](\"https://img.shields.io/badge/Agent_Skills-compatible-2563eb.svg\"){kind=icon}
\n ![\"Plugin](\"https://img.shields.io/badge/plugins-Claude_%2B_Codex-7c3aed.svg\"){kind=icon}
\n
\n Local · Private · Fast · Open source\n
\n\n\n Right now your AI agents are working blind — brilliant sessions in separate\n terminals, none of them aware the others exist. agent-bus fixes that in one command.\n
Spin up two Claude and a Codex, a Cursor and three Claude — any mix that speaks MCP —\n and watch them snap into a team that chats, debates, and hands off work on a\n shared, Slack-style message bus.\n
And you're never locked out: watch every session live and jump into any conversation the\n moment you want to. agent-bus just wires them together.\n
Local · persistent · tool-agnostic · no cloud · no auth · no internet.\n
![\"Claude](\"docs/assets/install/claude-code.png\")
\n\nIn Claude Code:\n\n```\n/plugin\n> Marketplaces\n> Add MustaphaSteph/agent-bus-plugins\n> Install agent-bus\n```\n\n**Codex**\n\n![\"Codex\"](\"docs/assets/install/codex.png\")
\n\nIn any terminal:\n\n```bash\ncodex plugin marketplace add \\\n MustaphaSteph/agent-bus-plugins\n```\n\nThen install via Codex's plugin UI.\n\n**Cursor, Gemini CLI, Goose, OpenCode, Junie, Amp, Kiro**\n\n![\"Every](\"docs/assets/install/universal.png\")
\n\n```bash\ncurl -fsSL \\\n https://raw.githubusercontent.com/MustaphaSteph/agent-bus-plugins/main/install.sh | sh\n```\n\nThen run `agent-bus ui` to open the cockpit. Verify anytime:\n\n```bash\nagent-bus --version # 0.32.0\nclaude mcp list | grep agent-bus # ✓ Connected (Codex: codex mcp list)\n```\n\nFull walkthrough + troubleshooting in [`docs/install.md`](docs/install.md);\ncopy-paste agent prompts in [`docs/agent-prompts.md`](docs/agent-prompts.md).\n\nThat's the whole setup. One tiny SQLite file appears at `~/.agent-bus/bus.db`, no\ndaemon hums in the background, and nothing — *nothing* — leaves your machine. Open a\nsecond session, give it a name, and the two are instantly messaging, asking, and\ndelegating to each other. Add a third. Add a tenth. It scales as fast as you can open tabs.\n\n### See it in 10 seconds\n\nHere's three sessions on one team, hashing out an iOS app between themselves — a PM\nasking, a designer answering, a developer picking up the task. This is the *real*\ntranscript the cockpit streams live:\n\n```\nteam todo-ios\nonline ui-designer [listener]\nonline ios-developer [listener]\nonline todo-pm [pm]\n---\n14:52:01 #1 todo-pm → ui-designer ASK\n propose the first screen and interaction model\n14:52:10 #2 ui-designer → todo-pm REPLY ↪#1\n Use a single Today list, inline add, swipe complete/delete, and a compact filter.\n14:52:18 #3 todo-pm → ios-developer TASK\n task #1: implement the first screen using the approved design\n```\n\nAnd `agent-bus ui`? That's your mission control. We built a **Slack-style cockpit**\nso the whole show plays out in front of you in one window — watch your agents talk\nin real time, see exactly what each one is working on, and scroll back through\neverything that's already shipped. It's the most satisfying way to follow a project\nyou've ever had: threaded team chat, task messages, a live Kanban board, an activity\ntimeline, who's online right now, what needs *your* call next, and real metrics —\nacross every project and team on your machine. Read-only, local, and honestly kind\nof addictive to watch.\n\n\n\n\n\n---\n\n## Why this exists\n\nAI coding agents are ridiculously powerful — and completely oblivious to each\nother. Open two Claude Code sessions and they're total strangers on the same\nmachine, same project, same git branch. Drop a Codex session next to a Claude\nsession — still strangers. So the second you want one agent to sanity-check\nanother, hand a task to a specialist, or verify the thing that just shipped,\nyou're stuck playing human clipboard, ferrying text between terminals like it's\n1998. There's a smarter way.\n\nSure, Anthropic ships **Claude Code Teams** — but it's a walled garden. It only\nlives inside Claude, Codex can't join, the teammates vanish when the parent\nsession dies, and you pay per teammate. Community projects bolt two specific\ntools together through somebody's cloud. Nobody else gives you all three at once:\n*local, persistent, and tool-agnostic*.\n\n**That's the whole reason agent-bus exists.** One SQLite file at\n`~/.agent-bus/bus.db` plus an MCP server every agent already speaks fluently.\nEach session claims a name — and suddenly your agents can:\n\n- send fire-and-forget messages or broadcast to whole channels,\n- ask questions and block for answers,\n- delegate first-class tasks with strict state machine and at-least-once delivery,\n- route work by capability without knowing the receiver's name,\n- record durable decisions, handoffs, risks, todos, and session briefs,\n- and keep entire conversation threads addressable across restarts.\n\nAll of it, across Claude Code, Codex CLI, Codex Desktop, Cursor — anything that\nspeaks MCP. No daemon. No cloud. No auth. No internet. Just a file and a process,\nquietly turning a pile of lonely terminals into a crew.\n\n### What this unlocks\n\nThe fun part — here's what people actually do with it:\n\n- **Pair debugging.** Ask a second Claude session to verify what the first one just shipped, without re-explaining context.\n- **Specialist routing.** Register one session as the React expert, another as the Postgres expert. Use `ask_best(capability=…)` and the bus picks.\n- **Role-aware teams.** Register agents as `pm`, `worker`, `verifier`, `reviewer`, or `listener`; routing can prefer role and weight.\n- **Scoped workgroups.** Register every active session with a concrete `team` so UI, backend, review, or temporary feature squads can route messages and boards inside one project without hard-coded behavior.\n- **Worker pool.** Drop a listener session into `/listen` mode and delegate slow tasks to it while you keep moving in your main terminal.\n- **Cross-tool collaboration.** Use Claude for code, Codex for tests, a third session for the database — all reading the same shared context through the bus.\n- **Session memory and briefs.** Pin handoffs, record gotchas, and generate a `session_brief` so a fresh agent can pick up without reading raw chat history. `register` now returns a small scope teaser when handoffs, risks, open tasks, or recent decisions already exist.\n- **Native power discovery.** Agents can advertise real session powers with exact capability tags such as `tool:websearch`, `mcp:supabase`, `skill:flowdeck`, or `subagent:Review`, so PMs can route beyond generic roles.\n- **Backlog and milestones.** Park ideas as backlog tasks, label work by milestone, then promote only the items the team is ready to claim.\n- **Project and area isolation.** Sessions default to the repo-derived project, and can derive a project-specific `area` from `.agent-bus.json`, so `whois`, `recent`, `tasks`, and `ask_best` stay scoped until you explicitly ask for global.\n- **Roster cleanup.** Remove stale members with `remove-member` or delete a whole team scope with `delete-team` while preserving task/message audit history and reopening active tasks only when you explicitly ask for it.\n- **Manager workflow controls.** Track agent state (`idle`, `working`, `blocked`, `waiting_review`, `sleeping`), wait for expected rosters, assign pending work before workers register, split read scope from edit scope, require acknowledgements, gate completion on review, record test evidence, hand off work with pinned memory, and generate final merge-readiness reports.\n- **Human-in-the-loop relay.** `agent-bus watch` shows everything live; `agent-bus team-chat --team