Free rendering
Install Chromium once after npm ci if you want Playwright rendering:
npx playwright install chromiumThe free crawler follows same-site links and sitemaps, respects access restrictions, uses a fresh browser context and does not reuse authenticated sessions. A page cap and crawl time budget keep runs bounded.
Optional server-side keys
Add FIRECRAWL_API_KEY for API scraping. GOOGLE_SEARCH_API_KEY and GOOGLE_SEARCH_ENGINE_ID can assist discovery for eligible existing customers. Automatic mode falls back to the free crawler when an API scrape fails. API-only mode requires Firecrawl and does not use HTTP fallback. Local robots checks apply in every mode. Configure AI_PROVIDER and its chat credentials separately for profile analysis after retrieval. Free crawler avoids search/crawl API calls but can still call the configured chat provider. Vision/video names remain reserved. NVIDIA_EMBEDDING_MODEL defaults to nvidia/nemotron-3.5-embed-1b; Azure uses AZURE_EMBEDDING_DEPLOYMENT. Persisted research and indexing require the worker.
Coverage is not completeness
Some directories block automated access or require accounts. Directory facts are marked unverified. A source report includes attempted URLs, status and failures; review those before using a suggestion in the blueprint.
Security
Research mutations require the company owner session, enforce same-origin requests and rate-limit runs. Private or reserved destination addresses are rejected and redirects are revalidated. A paid crawler is a separate external-service boundary.