lerp
A small, reliable CLI that orchestrates software work through Linear.
go install github.com/mattwalters/lerp/cmd/lerp@latestYou put tickets on a board; lerp runs coding agents to move them across it.
A recording, against a fake board and a stub agent.
One binary and one screen. Lerp opens on the Linear board you already have, fills a few lanes with coding agents, and answers the two questions an operator actually has: what is waiting on me, and what is the machine doing right now. Nothing runs on a server and nothing is stored anywhere but Linear — close the laptop and the work stops; open lerp again and the loop picks the same board back up.
The mental model
Linear is the database: all durable state — what work exists, what stage it is in, who has claimed it, what was decided — lives in Linear, and lerp keeps no store of its own. The board is the workflow: a queue is a Linear status with a prompt, a runner, and a status to move to on success. Lerp is a reconciler: desired state is the board, actual state is the agent processes running on this machine, and the loop starts, adopts, or reaps agents until the two match.
Start here
Overview — what lerp is in a page, and where the rest of the manual is.
Why lerp — what it deliberately is not, and how that differs from the other ways to put agents on a backlog. Read this one first if you are still deciding.
Quickstart
— lerp init, a first run, and a first
promoted ticket, following only the manual.
SCOPE.md — the fence around the project: five concepts, nine invariants, and the litmus tests every change runs through.
This page isn’t versioned — the manual is. Open Docs and the picker in its header moves between releases.