The checked command line

Needs Node.js 20 or newer. Install it globally, or run it with npx without installing.

npm install -g @checked/cli

npx @checked/cli --help

On your own machine, checked login shows a short code and opens the browser. Check the code matches, approve, and the terminal is signed in as you for 90 days. The sign-in is saved in your user config directory, readable only by you. checked logout revokes it on the server and removes it.

checked login
checked whoami
checked projects list

On a server or in CI, use a workspace API key instead: create one under Settings, API keys and set it as CHECKED_API_KEY. A key belongs to one workspace and can read results, import specs and start runs there; it cannot change workspace settings. To save a key on a machine, pipe it in: echo "$KEY" | checked login --with-key.

Projects and runs take their ref (as in the app’s URLs, for example payments-kqzxm and hbtqwpmz) or their number. Every command takes --json for scripts; results go to stdout and progress to stderr, so JSON output stays clean. Run checked <command> --help for the options.

checked import openapi.yaml --project your-project-ref
checked run --project your-project-ref --env Staging --depth critical
checked findings list --project your-project-ref --severity high
checked findings show r123
checked runs report <run-ref>
checked scan https://example.com
checked usage

checked import takes a file or a URL: OpenAPI 3, Swagger 2, Postman, Insomnia, HAR, GraphQL (a live endpoint is read by introspection), cURL commands or a plain list of endpoints. It shows what it found (format, endpoint count, how many fit your plan, a sample) and asks before saving. In CI, pass --yes; --dry-run only looks. Endpoints merge by method and path, so test history stays linked; --replace swaps the whole list.

checked run starts a run on the project’s default environment (or --env) and follows it live. API projects run their test suites; --depth critical runs the most important tests only, which is quicker and uses less allowance. Website projects run a health scan by default; --kind explore starts the AI explorer where your plan includes it.

When the run ends, the CLI evaluates the project’s release gates (pass rate, critical and high defects, and any others you set in the project) and exits with a code your pipeline can use. Defect counts can change when the AI report lands a minute later; --wait-report 2m waits for it before judging. --no-gates uses the run’s own pass or fail instead. --no-wait starts the run and returns at once with its ref.

# .github/workflows/checked.yml
name: Check’d
on: [push]
jobs:
  checked:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/setup-node@v4
        with:
          node-version: 20
      - name: Run the Check’d tests
        env:
          CHECKED_API_KEY: ${{ secrets.CHECKED_API_KEY }}
        run: npx -y @checked/cli run --project your-project-ref --depth critical --timeout 20m

# .gitlab-ci.yml (set CHECKED_API_KEY as a masked CI/CD variable)
checked:
  image: node:20
  script:
    - npx -y @checked/cli run --project your-project-ref --depth critical --timeout 20m

The run keeps going on the server if the job is cancelled or times out; open it with checked open <run-ref>.

Starting runs, scans and re-runs uses your workspace’s allowance, the same as in the app. When the current session’s allowance is used up, commands that start work stop with exit code 4 and say when it resets, for example: You have used this session’s allowance on Starter. It refreshes in 3 hours. Reading results, reports and findings is free. checked usage shows where you stand.

The MCP server gives Claude Code, Cursor and VS Code the same abilities as tools. See AI assistants (MCP).