Flightbox developer docs

Flightbox tests whether AI agents can complete real tasks on your website — across the agent matrix, fully recorded, scored, and explained.

• Quickstart — first test in five minutes
• Suites & tasks — the YAML schema and oracles
• The agent matrix — adapters, access methods, requirements
• REST API — every endpoint (interactive reference at /docs)
• SDKs — Python and JavaScript
• CI integrations — GitHub Actions, GitLab, anything with a shell
• MCP server — let your agents drive Flightbox
• Webhooks — batch notifications
• Self-hosting — Docker, compose, Render

Philosophy: metered, no feature gates, your spend caps. Runs that fail on our infrastructure are never billed.