provar/cli
The command-line interface. Compile and run .test.yml graphs from a terminal,
without spinning up the desktop editor. Three subcommands: setup scaffolds a new project (with --sample for the bundled
starter), compile turns graphs into machine-friendly scripts, run executes them against a target.
Quickstart
1. Install
curl -fsSL https://provar.se/install.sh | bash macOS and Linux. Windows users grab a binary from the GitHub releases page.
2. Scaffold a project
provar setup my-app --sample
cd my-app --sample gives you a working login file pointing at a live demo.
3. Add your API key
Edit ~/.provar/settings.yml. The CLI writes defaults on first run. Add an apiKey for the provider you want:
# ~/.provar/settings.yml
models:
provider: openai
providers:
openai:
model: gpt-5.5
apiKey: sk-...
google:
model: gemini-3.5-flash
anthropic:
model: claude-5-sonnet-latest Providers: openai, anthropic, google. Set a custom baseUrl on a provider entry to point at any OpenAI-shape endpoint (local Llama,
Ollama, etc.).
4. Compile and run
provar compile .
provar run . 5. Point it at your own app
Edit .provar/config.yml, change variables.baseUrl, recompile. When
the UI drifts, recompile the file and the agent rewrites the broken action.
Variables are overridable at run time via env: BASE_URL=http://staging.example.com provar run .
CI
# .github/workflows/provar.yml
name: provar
on: [push, pull_request]
jobs:
test:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v4
- run: curl -fsSL https://provar.se/install.sh | bash
- run: provar run . Pre-job secrets, matrix setups, GitLab / CircleCI recipes — see CI integration.