zlatko-lakisic

Agent Coop — AI skill for Claude Code

AI community

Let AI agents on your computer work together through plain files.

How to install Agent Coop

This entry records only its repository, not the path inside it, so there is no exact command to give. Open zlatko-lakisic/agent-coop and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agent Coop does

Let AI agents on your computer work together through plain files. One agent drives each project and the others answer, review, and help. Works with Cursor, Claude Code, Codex, Gemini CLI, Ollama, and LM Studio.

Alternatives in AI

BrowserSkill — Let AI agents use your real, logged-in browser without interrupting your work 9.1k ★
Claude Hub — by Claude Did This - A webhook service that connects Claude Code to GitHub repositories, enabling AI-powered c 393 ★
CLI Anything Zotero — CLI server for Zotero 7/8/9 — let AI manage your library 129 ★

README

Agent Co-op

Version 0.2.0. Early, so expect changes.

AI agents on one computer work together through files in a shared folder, with one driver per project and helpers that answer it.

The protocol is files, the skill is a prompt, and everything else is optional.

Protocol. Messages, roles, and folders. Described in SPEC.md.
Skill. A prompt each agent follows. No plugin and no server.
agentcoop command. It sends messages safely, installs the skill, and checks the setup.

flowchart TB
  driver["Driver
one per project"]

  driver --> codingAgents
  driver --> desktopChat
  driver --> localModels

  subgraph codingAgents["Coding agents"]
    direction TB
    cursor["Cursor"]
    claude["Claude Code"]
    codex["Codex CLI"]
    chatgpt["ChatGPT desktop
Codex"]
    gemini["Gemini CLI"]
    copilot["GitHub Copilot
planned"]
    opencode["OpenCode
planned"]
    cline["Cline
planned"]
    goose["Goose
planned"]
  end

  subgraph desktopChat["Desktop chat"]
    direction TB
    work["ChatGPT Work
planned"]
    desktop["Claude desktop
planned"]
  end

  subgraph localModels["Local models"]
    direction TB
    ollama["Ollama"]
    lmstudio["LM Studio"]
  end

  classDef role fill:#0f172a,stroke:#0f172a,color:#f8fafc
  classDef helper fill:#eef2ff,stroke:#4f46e5,color:#1e1b4b
  classDef soon fill:#f8fafc,stroke:#94a3b8,color:#475569,stroke-dasharray:5 4

  class driver role
  class cursor,claude,codex,chatgpt,gemini,ollama,lmstudio helper
  class copilot,opencode,cline,goose,work,desktop soon

A message is one JSON file. The sender writes it once. Nobody edi