What Hive is
Hive runs jobs — CI/CD pipelines, test suites, builds, ad-hoc work — as Kubernetes Jobs across a fleet of clusters. A run is a DAG of tasks; each task executes in a container on whichever runner the scheduler picks.
You define pipelines in Starlark (.hive/main.star in your repo), not YAML. Starlark is a real language with functions and conditionals, so a pipeline can compute its own shape — which is how changed-file-aware pruning works: the DAG for a docs-only commit is smaller than the DAG for a migration.
What it is not
It is not a replacement for your coding agent. Hive runs *alongside* Claude Code, Cursor, Codex, and GitHub Actions — it is the thing that executes and gates the work, not the thing that writes it. Repositories commonly keep their existing Actions workflows while adopting Hive for the parts that need a real fleet.
The pieces you will meet
| Thing | What it is |
|---|---|
| Project | One repository, registered with Hive |
| Pipeline | A named DAG defined in .hive/main.star — e.g. ci |
| Run | One execution of a pipeline, at one commit |
| Task | One step of a run, one container |
| Tenant | Your organisation. Everything above belongs to exactly one |
| Runner / cluster | Where tasks actually execute |
Tenancy, briefly
Your tenant is the outermost boundary. Projects, runs, users, clusters, secrets and credentials all carry a tenant, and a token issued to your tenant sees only your tenant's rows — another tenant's run id answers 404, not 403, because the existence of their data is not yours to learn.
See security-and-isolation for what that guarantees in detail.
This page is the hive-docs knowledge collection your agent reads inside the product, published verbatim. Markdown version