Agent Dispatcher plugin
A Claude Code and Codex plugin that finds task code by symbols and Git history, builds context and picks a specialist role.
- 58 ★ on the repository
- Last commit
- MIT
Extended access
Why this level
- Roles and recipes steer the agent toward code edits and running checks in the project.
- Optional model features may send code excerpts to the chosen provider.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/agent-dispatcherDetects 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.
- Roles and recipes steer the agent toward code edits and running checks in the project.
- Optional model features may send code excerpts to the chosen provider.
This is third-party code. Review the repository files before installing.
What it does
Agent Dispatcher finds the code a task needs using symbols, source text, relationships and Git history. It builds a bounded context packet with code excerpts and the reasons each file was picked, and matches work to a role: 27 roles cover engineering, design, testing, research, writing and operations. Small edits go direct, while larger tasks get specialist guidance. Checks become explicit: receipts distinguish passed checks, failures, zero tests and unverified work. The pack ships 79 local skills, 31 external skills, 8 recipes and 19 MCP servers with documented fallbacks. Standard retrieval runs locally with no model or network calls. Commands: `$agent-dispatcher` in Codex and `/agent-dispatcher:agent-dispatcher` in Claude Code, with controls such as `context build`, `doctor`, `health` and `status`.
Who it is for. Developers in large repositories where an agent spends context hunting for files and forgets to verify results.
Good fit when
- You need to find why checkout fails after login without overrunning context.
- The agent edits blindly and you want to see why files were chosen.
- A big task should go to the right specialist: reviewer, debugger, designer.
- The result needs an explicit summary of what passed, failed or was not tested.
Not a fit when
- You need native Windows installation: only macOS, Linux and WSL are supported.
- Ordinary repository search is enough without roles and recipes.
- Model-assisted features are on but source code cannot go to the provider.
Example request
Through the dispatcher with the debugger role find why checkout fails after login, fix the cause and show the verification summary.Limitations
It needs an installed, authenticated Claude Code or Codex, Git and Python 3.10 or newer. macOS, Linux and WSL are supported, and native Windows installation is not covered. Roles only tell the agent how to work: the model, tools and permissions come from the host. Automatic activation is opt-in, and the Codex hook must be reviewed and trusted through `/hooks`. Model-assisted features such as reranking and file summaries are off by default and, when enabled, may send the task and source excerpts to the chosen provider. Use one installation method per host, otherwise commands and hooks are duplicated.
How to disable. For Codex remove the skill in `~/.agents/skills/agent-dispatcher/` following `adapters/codex/README.md`. In Claude Code remove the plugin with `claude plugin uninstall agent-dispatcher@agent-dispatcher`.
FAQ
How do I invoke it in Codex and Claude Code?
In Codex: `$agent-dispatcher` plus the request. In Claude Code: `/agent-dispatcher:agent-dispatcher` plus the request; pick a role with a command such as `/agent-dispatcher:agent-reviewer`.
How can I see why a file was chosen?
`context explain` shows the role, skills and context selection, and `context build <request>` assembles passages without executing the request.
What does the doctor command check?
Installation health and setup recommendations. `health` summarizes capabilities and `inventory setup` shows where setup is missing.
Is an API key needed?
Not for normal use. Model-assisted features such as candidate reranking need your configuration and are off by default.
Does installation start background processes?
No, installing the pack connects no external services and starts no background agents. Automatic activation is enabled separately with `on here`.
Discussion
No questions or reviews yet.
Related
PluginsA Claude Code and Codex plugin with skills for code research, tests, refactoring and independent review of finished changes
Claude Code kit with eight phase agents, 21 skills and hooks that keep the agent inside set limits.
agentplay turns local Claude Code, Codex and Cursor sessions into a file map and a replayable step timeline.
Claude Code Skills and Agents
Claude Code Skills & Agents
A Claude Code plugin bundling skills and sub-agents: deep research, image generation, GitHub work, code review and model switching
A workflow template for Claude Code and five more agents: a backlog task, a diff review and a commit.
A plugin for Codex and Claude Code that runs a task through a chain of roles from planning to an independent audit
README in short
A pack of roles and context retrieval for Claude Code and Codex: 27 roles, 79 skills, recipes and explicit verification summaries. Base retrieval runs locally.