.ankra/pipeline.yaml, says what to run. Ankra writes it for you, runs it on every push and pull request, reports back to GitHub, GitLab or Bitbucket, and publishes only images it has scanned and gated.
Get a pipeline in three commands
1
Point a CI cluster at pipelines
An organisation administrator chooses the cluster and turns on its pipeline workers, once:
2
Ship a repository
From the repository’s checkout:Ankra analyses the code and opens a setup pull request with a generated
.ankra/pipeline.yaml. The next push already builds. For a repository with nothing to deploy, connect it instead with ankra pipeline repositories connect and commit your own file. See Examples.3
Watch it
The section
Why Ankra Pipelines
What you care about in CI, and how each piece is delivered.
How it works
From push to green check, in diagrams: planning, dispatch, authority, the digest chain.
Generated pipelines
How Ankra writes your pipeline, and how to convert an existing one.
Fast CI
Affected-only stages, the dependency graph, newest commit wins, build once.
Caching
Keys, restore, trust scopes, what Ankra generates, and reading a hit or a miss.
Merge train
Test every pull request as it will merge, several at once, then merge it.
Examples
Complete pipelines for services, monorepos, databases, libraries, nightlies and releases.
AI in pull requests
The review that reads your pipeline, merge when green, fixes for failed builds, previews.
AI Gateway
Slack, Teams, GitHub, GitLab and Bitbucket, with Ask and Agent modes.
CLI and MCP
The everyday commands, scripting with exit codes, and the MCP tools.
With coding agents
Claude Code, Codex and the Claude app: connect, install skills, close the loop.
Full guide and reference
Every setting, limit, diagnosis code and key, for when you need the detail.