Schema.org for AI agents — the open standard for how businesses describe their services to AI agents
Server has 7 tools with clear verb-noun naming (get_, search_, register_, validate_, fetch_). All tools have descriptions (avg ~150 chars, within baseline 34-392). However, input schemas are present but minimal: most parameters lack type constraints, enums, or detailed format guidance. Output schemas are not documented. Error handling is basic (returns error dicts but lacks recovery guidance or categorization). No tool annotations (readOnlyHint/destructiveHint). The register_business tool modifies state but description doesn't explicitly warn of side effects. Overall: functional definitions but missing production-grade rigor around schemas, output documentation, and error guidance.
Fetch an Agent Interface spec from a remote URL. Businesses can host their spec at a well-known URL (e.g. example.com/.well-known/agent-interface.json). This tool fetches and validates it.
Get all capabilities of a registered business. Returns the full spec including all available actions an agent can take with this business.
Get a complete example Agent Interface spec. Shows a fully filled-out spec for a restaurant with menu browsing, ordering, and order tracking capabilities. Use this as a reference when creating your own spec.
Get a blank Agent Interface spec template for a business. Returns a template that businesses can fill out to make their services accessible to AI agents. Like Schema.org markup but for AI agents.
Register a business interface spec in the local directory. Makes the business discoverable by agents through search_businesses. The spec is stored locally and persists between sessions.
Output schemas not documented. Tools return dicts but LLMs cannot infer field types, required vs optional fields, or nested structure. Forces LLMs to guess at response shape.
Input parameters lack type constraints and enums. 'category' and 'query' are free-form strings with no validation guidance. 'spec_json' accepts any string, no format hint or validation rule stated in description.
register_business modifies state (INSERT OR REPLACE into DB) but description does not explicitly state this is a write operation with side effects. Agents need to know which calls are safe to retry.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
Search for agent-accessible businesses. Find businesses that have published Agent Interface specs, making their services available to AI agents.
Validate an Agent Interface spec for correctness. Checks that all required fields are present and properly formatted. Returns a list of errors if the spec is invalid.
Error handling lacks recovery guidance. Errors return bare dicts like {'error': 'Invalid JSON'} or {'error': 'Business not found'} without suggesting next steps (e.g., 'Try search_businesses() first').
No tool annotations. Tools lack readOnlyHint (for get_*, search_*, validate_*) and destructiveHint (for register_business). Agents cannot infer safety properties without explicit hints.