Claude Agent Bus — Data skill for Claude Code
Claude Code plugin: parallel sessions on one machine see each other, take turns on shared services (dev server, bundler, simulator, database), and exchange messages — enforced in hooks, not documentat.
How to install Claude Agent Bus
This entry records only its repository, not the path inside it, so there is no
exact command to give. Open akaribrahim/claude-agent-bus and copy the folder into
~/.claude/skills/, or the file into ~/.claude/agents/.
What Claude Agent Bus does
Claude Code plugin: parallel sessions on one machine see each other, take turns on shared services (dev server, bundler, simulator, database), and exchange messages — enforced in hooks, not documentation.
Alternatives in Data
- Financial Services — Reference agents, skills, and data connectors for the financial-services workflows we see most — investment ba 34.5k ★
- Context Mode — Benchmark Results — Benchmarked against real outputs from popular Claude Code MCP servers, Skills, and dev tools 5.6k ★
- 12 Services Layer — Prompt 12: Wire Up Services Layer (Analytics, Policy, Settings, Sessions) 2.3k ★
README
agent-bus
A Claude Code plugin for running several sessions on one machine without them quietly ruining each other's work.
The problem
You open two Claude Code sessions, each in its own git worktree. Their file edits never collide — separate checkouts. Two other things collide constantly, and both do it silently.
**The shared service lies about whose code it is running.** A reloading dev server watches the tree it was started in. You edit your worktree while the other one owns `:3000`, and your change never runs: the endpoint keeps its old behaviour and you conclude your fix does not work. Or your test suite goes green — against somebody else's branch. Nothing announces this. The port hides which checkout is answering.
**Agents are helpful at the wrong moment.** The other session is halfway through editing `api/service.py`. Yours runs the build, sees a syntax error in a file it has never touched, and fixes it. Now two sessions are editing one file, each undoing the other, and neither knows why.
A README explaining the rules does not fix either one, because an agent has to remember to read the README. This does, because the decision is made in a hook: the command does not run, and the agent is told what is wrong and how to fix it.
What it looks like
Not invented — copied from a run. A session in `wt2` curls a dev server that another session started from a different checkout:
agent-bus: BLOCKED. "web" (the demo server on :8099) is serving a different
checkout, so this command would exercise that code and report it as yours.
serving : …/repo
you are in: …/wt2
started by: repo-main, just now
It serves the tree it was started in, so a request to it answers with THAT
checkout's files.
Point it at your worktree — this stops the current one and starts
yours, and tells the other sessions it moved:
agentbus serve web
Or run one command against your own tree and leave it there:
agentbus run web --
A session runs the
Related Skills
AI Crew Sync
Coordination bus for teams of AI coding agents — Claude Code, Codex, Cursor or any MCP client — sharing messag
Tfl Skill
Real-time London transit data for your OpenClaw agent. Get Tube arrivals, bus predictions, line status, disrup
Sdlc Dev
Phase 3 — Development & Coding. Spawns 9 agents with 4 parallel groups: backend-engineer, frontend+mobile, dat
Dibs
Call dibs on files. Coordination for parallel coding agents — file claims with expiry, enforcement hooks, and
Council Sim
AI-powered city council debate simulator for data center proposals. 5 AI agents debate a data center proposal
Curation Discipline
Keep an AI agent from silently corrupting your dataset: a staging gate agents cannot bypass, read-only enforce
Related Agents
Spec Miner
Extracts behavioral specs from an existing codebase into flat Requirement/Invariant blocks with machine-parsea
Agent Native Reviewer
Use this agent when reviewing code to ensure features are agent-native - that any action a user can take, an a
Comment Analyzer Readonly
READ-ONLY comment analyzer for Phase 1 pipeline. Verifies comments and commit messages against code. Cannot Ed