akeneo.com is Poor to agents.
Discry independently scored how well an AI agent can discover and understand the akeneo.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 · 24/100Comprehension
55% of score · 88/100What we found
- An agent that already has the docs URL would comprehend this API quickly — every error code (400/401/403/404/405/406/409/413/415/422/429) is documented with a realistic JSON body AND explicit recovery guidance, including 429 Retry-After handling and _links.documentation pointers back to the relevant page. This is best-in-class error-recovery documentation.
- An agent trying to DISCOVER this API would struggle: the docs host (api.akeneo.com, a static Google Cloud Storage site) returns 404 for robots.txt, llms.txt, llms-full.txt, sitemap.xml, and .well-known/mcp.json. There are no agent-oriented discovery signals at all.
- The one discovery win is a real, publicly downloadable OpenAPI 3.1 specification (JSON/YAML), auto-generating a Postman collection — an agent that finds it gets a complete, current machine-readable contract.
- Capability boundaries are explicit and agent-friendly: max 100 items per batch, a 10,000-record offset ceiling that returns a 422 telling the agent to switch to search-after pagination, and a 1,000,000-character-per-line limit — an agent learns the limits from docs rather than by failing.
- No AGENTS.md exists in the akeneo/pim-api-docs repo and no Akeneo MCP server is listed in the major registries (Glama, Smithery, PulseMCP), so an agent has no coding-agent or tool-calling entry point beyond the raw REST spec.
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 api.akeneo.com covering the core connector flow (create connection → generate token → fetch/patch products, with links to auth, pagination, and the OpenAPI spec). This is the single highest-impact discovery fix and would lift the weakest dimension.
- 02Add a robots.txt and sitemap.xml to the GCS-hosted docs site that explicitly allow AI crawlers and list the documentation pages — right now all four discovery files 404, which suppresses agent and search indexing.
- 03Add an AGENTS.md to the akeneo/pim-api-docs repository summarizing the auth flow and endpoint scope so coding agents get useful context without parsing the full reference.
- 04Publish or register an official Akeneo MCP server (and list it on Glama/Smithery/PulseMCP) to give agents a first-class tool-calling path into the PIM.
- 05Surface multi-language request examples (Python, PHP, JS) alongside the curl snippets in the guides — the values are already realistic and copy-pasteable, but curl-only narrative examples limit direct reuse by agents targeting other runtimes.
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 unusually well: OAuth2 password-grant flow with refresh tokens, a machine-parseable JSON error format (code/message plus a structured errors array), 429 Too Many Requests with a Retry-After header, and two fully documented pagination methods (search-after cursor for high-volume entities, offset/page as default). Idempotency keys are not documented.