Start it once
On a clone you never develop in, run lumpcode start. It discovers every lump in the repo and works through them on a schedule.
Tutorial
In the morning, branches are waiting. You merge, you push a new lump, the worker picks it up. Nothing to deploy.
On a clone you never develop in, run lumpcode start. It discovers every lump in the repo and works through them on a schedule.
Nothing to deploy, register, or restart. The worker fetches your primary branch on its next pass and finds the new lump on its own.
It works through each lump context by context and pushes a branch per run. You open the pull requests and merge from wherever you are.
If you have not run a lump by hand yet, do that first: From install to the first PR. This page is only the worker.
Clone the same repo into a folder you never edit. Another directory on this laptop is enough, for example ~/lumpcode-worker. A small always-on box is later, not required.
.lumpcode/ is already in the repo, so do not run project-setup on the worker. Put the CLI on this clone, install project deps, write .lumpcode/local.json, then start. The worker also needs Node.js 22+, git origin fetch and push, and a CLI agent on PATH (already logged in).
{
"mode": "dedicated"
}mode: dedicated wipes uncommitted work on this clone. Do not pick it for a checkout you also edit.
npm install -g @lumpcode/cli
npm install
lumpcode start
lumpcode daemon-statuslumpcode start with no filters discovers every lump on a schedule. You do not restart for new lumps.
Later: lumpcode daemon-status, lumpcode daemon-log, lumpcode stop.
Writing TypeScript lumps? Lump config.
Author as in the first PR. Preview with lumpcode lump-plan <lumpName>. Commit, push, and merge to the primary branch. The worker fetches that branch on its next pass.
LUMP: <lumpName> - <context> marker commit on a lump/… branch. Open a PR. If you squash, keep that line. disabled: true on the lump, push, and merge. The next pass soft-skips it. The worker stays up. Depth: local.json, run vs start.