amadeus.com is Needs Work to agents.
Discry independently scored how well an AI agent can discover and understand the amadeus.com 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 · 52/100Comprehension
55% of score · 87/100What we found
- An agent that already knows the docs URL is well served: Amadeus publishes a public, MIT-licensed OpenAPI 3.0 spec per API on GitHub (amadeus4dev/amadeus-open-api-specification) with task-oriented field descriptions and realistic examples (IATA codes like SYD/BKK/CDG, ISO-8601 dates), plus clean mkdocs-markdown guides for auth, errors, rate limits, and pagination.
- An agent trying to DISCOVER the API natively would come up empty: there is no llms.txt, no llms-full.txt, no .well-known/mcp.json, and no AGENTS.md in the amadeus4dev org — every agent-native discovery signal returns 404 or redirects to the homepage.
- The error documentation is a standout for agent self-correction: each 4xx/5xx scenario lists the numeric code AND an explicit recovery action (401 expired → generate a new token; 429 → check rate limits and adjust; 404 → verify the endpoint/spelling), which is what separates agent-usable docs from human-only reference.
- The human-facing API reference pages are JS-rendered Swagger UI that scrape to near-empty markdown — an agent relying on the rendered docs (rather than the GitHub OpenAPI JSON) would lose almost all endpoint structure.
- Time-sensitive risk: a site banner states the Amadeus for Developers self-service portal will be decommissioned on July 17, 2026 (weeks away), after which these scanned self-service docs move behind the Enterprise portal — agent access to this documentation may change materially.
What to change
Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.
- 01Publish an llms.txt at developers.amadeus.com/llms.txt with a titled, API-focused summary and links to the core flows (auth, flight/hotel/transfer search & booking) — this single quick win lifts three discovery checks (llmsTxt, llmsTxtQuality, and token efficiency) that currently all fail.
- 02Add an AGENTS.md to the primary GitHub repos (amadeus-open-api-specification, developer-guides) pointing coding agents at the OpenAPI spec files and SDKs — cheap to add and a rising discovery standard.
- 03Serve the API reference as clean markdown (or content-negotiated text/markdown) instead of client-side-rendered Swagger UI, so agents scraping the reference pages recover endpoint method/URL/parameters rather than an empty shell.
- 04Expose a .well-known/mcp.json (and/or ship an official Amadeus MCP server) so agents can auto-discover tools and auth — today only third-party community MCP servers exist on Smithery/Glama/PulseMCP.
- 05Given the July 17, 2026 self-service decommission, ensure the equivalent OpenAPI specs, guides, and discovery files are mirrored on the Enterprise portal so agent-readiness does not regress after the migration.
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 characteristics are documented thoroughly for humans: OAuth2 client-credentials flow with 30-minute Bearer tokens, a machine-parseable JSON error envelope carrying numeric code/title/detail/status, explicit per-environment and per-API rate limits (10 tps test / 40 tps production), and a dedicated pagination guide. Idempotency keys are not surfaced for the self-service booking endpoints.