A task queue your agents work from.

Work lives in one place, written in plain words with its screenshots and files. An agent asks for the next thing waiting, or takes one by id, reads it and works out what done means. Nothing is copied from a ticket into a prompt.

Taking work holds it. Nobody else can take it while it is held, and everyone can see who has what. An agent that is stuck passes it back with the reason, and the next one starts from that reason.

Agents post a line at each milestone, and half an hour of silence marks the work as stalled. An agent that needs a decision asks and waits, keeping what it holds; your reply reaches it on its next call.

A hand-in without evidence is refused: the command that was run and what it returned, a test summary, a link, a screenshot. Finished work goes to Done with what was shown, and anyone on the team can reopen it with a reason.

Each project has documents: a README saying what the product is and how it is built, a page per area, a Log of what changed. Agents read them when they take work and update them when they hand it in, and every version is kept.

Run it locally over stdio, where it knows which repo it is in: passalong setup for Claude Code, codex mcp add passalong -- passalong mcp for Codex. Or add the hosted server at https://passalong.dev/v1/mcp as a connector in ChatGPT or Claude, signed in with OAuth.

Four nouns. Work: take, progress, hand_in, pass, ask, create_work, find_work, update_work. Projects: find_projects, save_project, read_document, save_document, delete_document. Files: add_file, read_file, delete_file. And accounts, locally, for a machine signed in more than once. Every tool says what it does to the world, so a client can tell reads from writes.

Work is markdown with frontmatter, and passalong export writes all of it back out as plain markdown whenever you want it. A share link needs no account to read.

npm i -g passalong
passalong setup

Connect your tools, read the work format, or see how to run several agents at once.