About Workhorse

What Workhorse is, who publishes it, what it promises, and where its source lives.

Workhorse is a durable job queue that lives inside PostgreSQL. An application enqueues a job in the same transaction that writes its own data, so the job and the data commit together or not at all. Workers in TypeScript, Python, and Go claim and run those jobs through one SQL protocol, and the database keeps the evidence: every attempt, checkpoint, wait, and outcome is a row an operator can query.

Who publishes it

Stablemates publishes Workhorse and maintains it in the open at github.com/stablemates/workhorse. The packages are released under the Apache License 2.0, and contributions are accepted under the contributor agreement in that repository. There is no paid tier today. The beta withholds nothing; a later commercial product would be new files under its own licence, not a second licence of the same code.

What ships

Three lines ship from the same repository and speak the same protocol:

  • The TypeScript package @stablemates/workhorse on npm, which also carries the schema tool, the worker runtime, the operator dashboard, and the workhorse command line.
  • The Python distribution stablemates-workhorse on PyPI.
  • The Go module github.com/stablemates/workhorse/go.

The current published version of each line, and which versions receive fixes, are on Releases. The tested runtimes and PostgreSQL versions are on Compatibility.

What it promises, and what it does not

Workhorse is a public beta. It is usable for evaluation and early production adoption. A minor release may change behaviour, so read the changelog before you upgrade. Inside a major line a migration only adds, so a running deployment upgrades in place and is never asked to recreate its database.

Handlers run at least once, PostgreSQL is the only supported database, and there is no separate workflow definition language: durable steps are composed in ordinary handler code. Limitations lists every boundary with the workaround where one exists, and Core concepts explains the model behind the guarantees.

Where to start

Quickstart installs the schema and runs a first job. Coding agents start at Workhorse for AI coding agents, and llms.txt indexes every page in a form an agent can read. Questions and bug reports go to the channels on Contact.