Use Krypton where your agent and your team already work
Coding agents
Any agent that can run a command can use Krypton. Add the rules from Getting started to CLAUDE.md, AGENTS.md or your tool's rules file. Use --json if a hook needs structured output.
Pre-commit and CI
krypton scan . --base origin/main || { echo "Krypton: structure check failed"; exit 1; }
A GitHub Actions job (build the client once and cache it, or keep a built krypton in your image):
name: Krypton
on: pull_request
jobs:
structure:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
with: { fetch-depth: 0 }
- run: docker build -t krypton path/to/client
- run: docker run --rm -v "$PWD:/repo" krypton scan . --base origin/${{ github.base_ref }}
Or use the Krypton action, which sets one Krypton gate line in each pull request's checks and writes the scorecard to the run summary and to one pull request comment it edits on every push, never a second one. See Review.
Docker
docker build -t krypton client
docker run --rm -v "$PWD:/repo" krypton scan .
The image has git in it, so --base, order and --log work on the mounted repository.
The hosted demo
The repository includes a small web app that shows what a codebase costs to read and what a refactor could save.
- Visitors sign in with GitHub. GitHub does the authenticating and decides which repositories they may read; Krypton keeps no accounts.
- The server reads the chosen repository in memory with the visitor's own access, builds the syntax trees and sends them to the Krypton server.
- The page shows files, lines, tokens per read, structure score and the hardest functions, then a savings view built from the real size.
The savings use the rates measured on one repository (see Prompt caching), scaled to the visitor's repository, so treat them as estimates. Deployment steps (Docker and Railway) are in dashboard/DEPLOY.md.