Landing explains CI failures, fixes delegated problems, triages issues, and publishes review feedback where your team works. It uses your existing tools and checks.

With Landing, you own and control the workflow. Choose the model, prepare the execution environment, and adapt instructions and skills to your project. Run it in CI, locally, or as a service. Landing is open source under Apache-2.0 and requires no hosted Landing account.

Start in CI

Set up your first PR review. Configure a model, add the GitHub Action to your prepared workflow, and open a PR. Landing publishes review feedback with code comments for actionable findings. Your team decides what to accept using that feedback and independent checks.

Configure CI -> Open a PR -> Read the feedback -> Decide what to accept

Set the LANDING_MODEL repository variable and LANDING_API_KEY secret, and grant the review job contents: read and pull-requests: write. After checkout and successful native checks, add this step to your existing workflow:

- uses: bubbuild/landing@main
  continue-on-error: true
  env:
    GH_TOKEN: ${{ github.token }}
    LANDING_MODEL: ${{ vars.LANDING_MODEL }}
    LANDING_API_KEY: ${{ secrets.LANDING_API_KEY }}
  with:
    command: review
    instruction: Native checks passed for this checkout. Review this PR using repository guidance and publish actionable findings inline.

The example follows main; pin a release tag or reviewed commit to keep the runtime fixed. continue-on-error keeps automatic feedback advisory; inspect the Action logs if no review appears. The Action uses your authenticated gh, project tools, and skills. It checks who can delegate before starting the agent. See CI integration for other CI systems and GitHub integration for /landing commands, publishing identities, and trust policy.

To build the same review and delegated-fix experience, ask your agent to read Landing's workflows and development guide, then adapt the CI setup to your repository's tools and checks.

Use it locally

With Python 3.12 or later, uv, and a POSIX host:

uv tool install landing
export LANDING_MODEL="openai:gpt-4.1"
export LANDING_API_KEY="your-provider-api-key"
landing explain "Explain this failure and the next useful check." --input check.log

Run from your project checkout and use a saved UTF-8 log for check.log. The commands are explain, triage, fix, review, and bump; each works independently. Local use shows how to delegate work and inspect the result.

Work with your team

Reliable results draw on your documentation, work items, infrastructure, tests, benchmarks, and observability. Working with Landing explains how to connect that context and retain feedback. Automate the engineering work you can delegate, and use the time saved to stay involved with contributors and users. Community Over Code.

Make Landing work for you covers project instructions, skills, and tools. For shared execution, run the server or deploy the container from ghcr.io/bubbuild/landing:latest. Actions and model history are stored in SQLite; workspace files stay in your workspace, and model requests go to your chosen provider.

Contributions are welcome. Licensed under Apache-2.0. Powered by Bub and tape.systems.