DEV Community

Vedant Shah
Vedant Shah

Posted on Edited on

Clawd HQ: turn your Claude Code session into a team of specialist agents

I kept hitting the same wall with Claude Code: one long chat doing everything in sequence. New requests queue behind a long job, and changing an instruction means stopping it and starting over.

So I built Clawd HQ, a free, MIT-licensed Claude Code mod. It turns your main chat into a chief of staff that hands work to specialist agents running in the background.

See the office

Clawd HQ office demo showing specialist agents

Office demo from the repository. This illustrates the panel, not a benchmark or evidence of faster work.

How it works

Delegates by default. Clawd hands every job of more than two or three steps to a specialist in the background. Jobs that don't depend on each other run side by side, and you can keep talking to the main chat while three specialists work.

Model and effort sized per job. The specialist doing the hard build can get a strong model; the one checking reports gets a small, fast one.

Specialists don't stall. They never stop mid-run to ask a question. They make their best call, finish the job, and list their assumptions and questions at the end for Clawd.

Steer a running specialist. It uses Claude Code's own agent messaging: Clawd can message any running specialist and it picks the message up at its next step. You can change or add to a brief without restarting the job.

A live office view. Start a session and a pixel-art office opens on its own, showing which specialist is doing what. Type /office-demo for a 90-second tour.

Install

Needs Claude Code v2.1.287 or later.

claude plugin marketplace add vedant-shah/clawd-hq
claude plugin install clawd-hq@clawd-hq
Enter fullscreen mode Exit fullscreen mode

Try one concrete workflow

Use a disposable project, or one you are comfortable letting Claude read. After installing, start a session and run /office-demo for the built-in 90-second tour. Then try a real, read-only brief:

Map this repository and find its existing test commands. Have specialists inspect independent parts in the background. Do not edit files, install packages, or run tests yet. Bring back one report with the paths you checked.

While the background work runs, ask a follow-up such as "Which test covers the login flow?" The thing to check is whether you can keep talking to the main chat while specialists work, and whether the final report is useful. This is a suggested trial, not a result I measured.

The plugin is free; model calls still use your Claude plan or API billing. It is a Claude Code mod, not a Cowork or Claude Chat extension. Early and still changing.

Repo: https://github.com/vedant-shah/clawd-hq

It's free and MIT-licensed. If you try it, I'd genuinely like to hear where it helps and where it gets in the way - issues and stars both welcome.

Top comments (0)