Documentation
Getting Started
Choose a route through Agent Surface by goal: make a product usable by agents, assess it, or build agents of your own.
Last verified 2026-09-25
Choose your path
| Goal | Start with | Then read |
|---|---|---|
| Let agents find your product | Discovery | Retrievability, Protocols |
| Let agents operate your service | API Surface | Tool Design, MCP Servers, Authentication |
| Make a repository easier for coding agents | Context Files | CLI Design, Error Handling |
| Make a web app usable by browser agents | Agentic UI | Discovery, Protocols |
| Let agents buy from you | Agentic Commerce | Product Data, Checkout Protocols, Payments |
| Prove agents can complete real tasks | Testing | Product Journeys |
| Assess where you stand | Scoring | Rubric, Public Scanners |
| Build an agent of your own | Agents | Agent Retrieval, Tooling Catalog |
If you are starting from zero
For an existing product, this order gets the most value earliest:
- Write accurate context files:
AGENTS.mdand any tool-specific files that import it. - Make content discoverable: policy,
llms.txt, structured data, and routes. - Make command output parseable: JSON, stable exit codes and strict input validation.
- Publish machine-readable contracts: , schema metadata and tool definitions.
- Add recovery paths: errors, retry guidance, and IDs.
- Secure programmatic access: scoped credentials and protected resource metadata.
- Expose tools deliberately: an server with a curated tool set, annotations and approval for risky actions.
- Test real tasks: cases, traces, CI gates and a before-and-after scorecard.