View as markdown

Hive documentation

Hive runs multi-step work as a folder-as-agent pipeline: each task moves through stage folders, an AI agent does the work at each stage, and a daemon advances it in the background, asking for answers only when it needs them. Its flagship coding workflow turns a rough idea into a merge-ready pull request — but coding is one of several workflows. Hive also ships content (research) and lets you author your own. These docs cover installing it, the ideas behind it, configuring it, the command surface, and running it as a service.

Start here

  • Getting started — install, hive init, and your first task in the TUI.
  • Concepts — folder-as-agent, the stage state machine, and the marker protocol.
  • Configuration — project config, patrol mode, reviewers, agent profiles, and the daemon.
  • Command reference — every user-facing hive command.
  • Custom workflows — the engine is generic: author your own per-project pipeline in YAML.
  • Operating — daemon, bot, and babysitter as services.

Looking for the source?

Hive is open source (MIT) on GitHub. The full engineering wiki lives in the repo; these docs are the curated, user-facing subset.