SheerID is Good to agents.
Discry independently scored how well an AI agent can discover and understand the SheerID API from what’s public — not whether it’s usable. Below: every signal we checked, what’s costing the score, and what to change.
SCORED UNDER RUBRIC 1.2 · A full re-launch under Discry Score 2.5 — a new behavioral instrument, not comparable to these scores — is in progress.
Discovery
45% of score · 71/100Comprehension
55% of score · 93/100What we found
- An agent landing on developer.sheerid.com gets a purpose-built map: the llms.txt explicitly addresses 'AI agents and LLMs', describes every section, and links the machine-readable OpenAPI spec — one of the strongest llms.txt files in the corpus.
- Error documentation is exceptional for self-correction: every common errorId has a recoverable/non-recoverable flag and a 'what to do' column, and the docs explicitly disambiguate the overloaded 429 status (only apiRateLimitExceeded is retryable) — an agent would avoid retry loops that most APIs' docs would cause.
- The REST API quickstart walks the full multi-step verification flow end-to-end (start → submit → success/docUpload/pending) with realistic request/response pairs, timeout guidance, and idempotency notes, so an agent can chain the whole workflow from one page.
- A public OpenAPI 3.0 spec (services.sheerid.com/rest/v2/swagger.yaml) is linked directly from llms.txt, plus a dedicated guide for generating typed clients from it.
- Agent-ecosystem discovery signals stop at the docs site: no MCP server or registry listing, no .well-known/mcp.json, no AGENTS.md in the SheerID GitHub org, and no llms-full.txt — an agent searching the MCP ecosystem would not find SheerID at all.
What to change
Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.
- 01Ship an official MCP server for verification operations and list it on Glama/Smithery/PulseMCP, plus publish .well-known/mcp.json — the largest remaining discovery gap for an otherwise agent-ready API.
- 02Add an llms-full.txt with the full docs content inlined so agents can ingest the complete Developer Center in one fetch.
- 03Add AGENTS.md to the primary SheerID GitHub repos (e.g. the JS library) with integration context for coding agents.
- 04Add multi-language code samples (curl/Node/Python) alongside the raw HTTP examples in the quickstart and endpoint reference — currently examples are HTTP/JSON only.
- 05Publish concrete rate-limit numbers (requests/sec and the Retry-After behavior) rather than only naming apiRateLimitExceeded.
Execution coverage · INFORMATIONAL, UNSCORED
Whether an agent can actually complete a call and recover from errors is the deeper Audit layer — documented here, but not part of the Discry Score.
Execution behavior is unusually well documented: machine-readable errorIds with an explicit per-429 retry table, an 11-second response-timeout recommendation, explicit non-idempotency of POST /verification with verificationId-reuse guidance, and webhook signature verification with retry/idempotent-handler notes. Specific numeric rate limits are not published, and the flow-based API has no documented pagination scheme.