JSM HQ CLI
JSM HQ runs a team of Claude Code and Codex agents on your repository and shows their work in a pixel-art office.
- 93 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- Agents write code, make commits and open pull requests in your repository.
- In Local mode agents run directly on your machine rather than in a sandbox.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/jsm-hqDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Agents write code, make commits and open pull requests in your repository.
- In Local mode agents run directly on your machine rather than in a sandbox.
This is third-party code. Review the repository files before installing.
What it does
JSM HQ opens in the browser and runs a project as a team. One agent, the lead, plans the work and hands out tasks, and each worker gets its own desk, its own copy of the repository and its own Docker Sandbox, writes code on a separate branch and opens a pull request. One task, one branch, one pull request. HQ follows pull requests on GitHub: review comments go back to the author, a passing pull request is reported to the lead and a merged one closes the task. Anything an agent needs from you, such as a decision, a key or an unblocked website, lands in the Action items list. A person always does the merging. Each agent runs Claude Code or Codex.
Who it is for. Developers who want to spread repository tasks across several agents and review only finished pull requests.
Good fit when
- A feature must be split into small tasks that several agents do in parallel.
- Each agent must work in an isolated sandbox without touching the rest of the system.
- You want to see the plan on a board, agents at desks and pull requests moving through review.
- CodeRabbit, Greptile or Copilot do the review and comments must go back to the author automatically.
Not a fit when
- The repository has no GitHub remote: each agent clones it, pushes a branch and opens a pull request.
- You have no Claude or ChatGPT account for the agents themselves.
- The task is small and one agent in a chat is enough.
Example request
Start HQ in this repository, pick Claude Code as lead and two Codex workers and ask the lead to plan two small tasks and assign the first one.Limitations
Node.js 20 or newer, git and a GitHub repository, GitHub CLI `gh`, Docker Sandboxes (`sbx`), tmux on macOS and Linux, and a Claude or ChatGPT account are required. Sandboxes run on macOS 14 or newer with Apple Silicon, Windows 11 and Ubuntu 24.04 or newer. On another system switch agents to Local mode in settings: they run Claude Code or Codex directly on your machine with permission prompts. After installing Docker Sandboxes run `sbx policy init balanced` once. Websites in the sandbox stay blocked until you allow them.
How to disable. Stop the team in the HQ panel and end the `jsm-hq` process. Remove the global package with `npm uninstall -g @therealjsmastery/jsm-hq` and delete branches and pull requests on GitHub.
FAQ
How do I start JSM HQ?
Inside a git repository with a GitHub remote run `npx @therealjsmastery/jsm-hq@latest`. The browser wizard then lists missing tools and the commands to install them.
Which agents work in the team?
A lead and workers, and each can be Claude Code or Codex. Agents message each other through HQ.
Does HQ merge changes itself?
No, merging stays with a person. HQ only follows pull requests, returns review comments and reports passing ones to the lead.
How do I install it globally?
Run `npm install -g @therealjsmastery/jsm-hq` and start `jsm-hq` in any repository. Flags are `--repo`, `--port` (default 7420) and `--no-open`.
What if a check shows an exclamation mark but the tool works?
You may have an old HQ cached by npx. Run `npx @therealjsmastery/jsm-hq@latest` to get a fresh one.
Discussion
No questions or reviews yet.
Related
CLI toolsJoxo puts a team's Claude Code, Codex and Cursor agents in one project with a shared task board and handoffs between sessions
The squads CLI runs teams of agents defined in markdown files in git: a lead briefs, workers act, a verifier checks.
A team of AI agents from one request: it plans the task, writes code in parallel branches, tests, reviews and opens a pull request
Mjolnir runs Claude Code, Codex and other agent sessions in a terminal, in containers and over SSH, and keeps them running.
DevChain runs a team of AI coding agents on your own machine with a shared board, chat, code review and live terminals.
A skill that turns several Claude Code, Codex and Cursor windows into a team with a coordinator, shared registry and cross-review
README in short
HQ lets you pick stack tools (Next.js, Clerk, Stripe, Supabase) and gives agents their official skills and docs access. The getting-started guide takes about 15 minutes.