Skip to main content

Quick Start

This is the intended standalone first-run flow from a source checkout, which includes the bundled demo tasks and fixture repository. The current public npm package is not the release path for this workflow.

1. Install Agent Harness

Follow the current install guide. The verified source path uses Node.js 20.x and runs commands through npm run harness --.

2. Generate config

npm run harness -- init --profile local

3. Validate setup

npm run harness -- doctor

4. Inspect the built-in demo

npm run harness -- demo --list-tasks
npm run harness -- demo --dry-run --task ts-isUUID

Dry-run before you spend model time

A dry run confirms the scenario and task shape without changing a repository or calling the model. Use it whenever you are trying a new example.

5. Run one task

npm run harness -- run tasks/ts-lib/ts-isUUID.json --repo test-repos/ts-lib

6. Inspect results

npm run harness -- status
npm run harness -- report

7. Recover from the common failure

If the model endpoint is not reachable:

  1. start Ollama or LM Studio
  2. rerun npm run harness -- doctor
  3. rerun the task

If you only want to validate local paths and config before services are up:

npm run harness -- doctor --skip-model-endpoint --skip-agent-brain