AI Agents Orchestrator plugin
AI-Agents-Orchestrator
The AI Collaboration plugin hands a task between Claude Code, Codex, Cursor, Gemini CLI and Copilot via a shared SQLite history.
- 101 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Hooks write prompts, commands, file paths and Git state into a local database.
- Secret fields and Bearer tokens are masked, while the rest of the content stays in the database.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/ai-agents-orchestratorDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Hooks write prompts, commands, file paths and Git state into a local database.
- Secret fields and Bearer tokens are masked, while the rest of the content stays in the database.
This is third-party code. Review the repository files before installing.
What it does
The repository holds six systems, but the `AI-Collaboration-Skills` folder installs on its own: a plugin for Claude Code, Codex, Cursor, Gemini CLI and GitHub Copilot CLI. Hooks write prompts, tool calls, commands, Git state and test results into a local SQLite database at `~/.ai-collaboration`, and the next agent receives context assembled from it without an LLM call. The bundle has five skills (resume-task, inspect-history, collaboration-status, create-checkpoint, handoff-task), an `ai-collaboration` CLI and an MCP server with ten tools. No Python packages are needed.
Who it is for. Developers who switch between several AI agents in one repository and lose the task context.
Good fit when
- Claude Code hits a limit and Codex continues the work in the same repository.
- You need to recall which tests ran in the last session and with what exit code.
- A checkpoint with the next step is saved before a break.
- A team passes a task between agents through the verified handoff-task skill.
Not a fit when
- Security policy forbids keeping prompt and command history on disk: hooks write it by default.
- You need a workflow orchestrator with roles: the plugin only passes context, while the Orchestrator and Agentic Team sit separately at the repository root.
- The environment has no Python 3.8 under the name python3.
Example request
Continue the task from the last session: use resume-task to show what is done, which tests failed and what the next step is.Limitations
It needs Python 3.8 or newer under the name python3 and Git to capture repository state. Hooks write the fields a given client provides, so event completeness differs between agents, and Cursor and Copilot get no hook before a tool call. The database is shared by all clients and lives at `~/.ai-collaboration`, and the `AI_COLLABORATION_HOME` variable changes the path. Secret fields and Bearer tokens are masked automatically, while the rest (commands, paths, Git state) stays in the database, so restrict access and clear history. The plugin installs from a local copy of the repository.
How to disable. Remove the ai-collaboration plugin in the client and the `~/.ai-collaboration` folder if the history is no longer needed.
FAQ
How do I install AI Collaboration in Claude Code?
Clone AI-Agents-Orchestrator and run `claude plugin marketplace add ./AI-Collaboration-Skills`, then `claude plugin install ai-collaboration@ai-collaboration`. Codex, Cursor, Gemini CLI and Copilot use different commands described in the plugin folder README.
Where does the plugin write data?
To `~/.ai-collaboration`: the `ai-collaboration.db` file and compressed objects in the `objects` directory. The `AI_COLLABORATION_HOME` variable changes the location.
Can a hook block the agent when writing fails?
No. A hook always exits with code zero and prints `{}`, so a write failure does not stop a tool call.
Which CLI commands exist?
They include `task start`, `status`, `context`, `sessions`, `search`, `checkpoint`, `task list` and `doctor`. Every command emits JSON.
Do I need to install the whole repository?
No. The plugin sits in the `AI-Collaboration-Skills` folder and imports nothing from `orchestrator/` or `agentic_team/`, so it installs on its own.
Discussion
No questions or reviews yet.
Related
PluginsA plugin for Claude Code and Codex: a shared task wall, project memory, reports and a dashboard that shows agents' work across sessions
AI Builder Club Skills
AI Builder Club — Skills
A Claude Code plugin marketplace to make a repo agent-ready and run compounding agent loops
A bridge between local CLI agents: Codex, Cursor or Claude Code delegate work to Antigravity, Grok, Kimi and seven more.
A skill and the ai-collab tool for Codex CLI, Claude Code and API models to work on a project, plus file reading in ChatGPT
A local dashboard for running many agents across machines: memory, code graph and agent-to-agent messaging
Shared infrastructure for Claude Code, Codex, Antigravity CLI and OpenCode: a task lifecycle, skills and sandboxes.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| AI Agents Orchestrator (this card) | no | 101 | |
| Table Skills: project-handoff | no | 463 | a single skill with a PROJECT_STATE.md file; this one has hooks, a SQLite session history, MCP and a CLI for five agents |
| Episodic Memory | no | 499 | searches past conversations by meaning; AI Collaboration tracks tasks, checkpoints and Git state |
README in short
The rest of the repository: the Orchestrator runs step-based workflows, Agentic Team runs a free dialogue between roles, Graphify builds a project graph, a 34-tool MCP server links them to IDEs and the Context Dashboard shows memory. The plugin needs none of these systems.