Epidemic Sound is Good to agents.
Discry independently scored how well an AI agent can discover and understand the Epidemic Sound 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 · 60/100Comprehension
55% of score · 100/100What we found
- An agent can bootstrap almost entirely on its own: a valid OpenAPI 3.0.1 spec is published at a stable public URL (partner-content-api.epidemicsound.com/docs/spec.json) and explicitly recommended for feeding to 'LLMs and AI coding assistants', alongside an official MCP server (public beta) and a dedicated 'Soundtracking with an LLM' guide — this is LLM-first documentation.
- Error recovery is best-in-class: the errors page maps every status code to a Retry?/backoff/'safe to show users?' decision and ships a copy-paste retry function, so an agent knows exactly what to do on 429/500/503 without failing first.
- Capability boundaries are stated up front — batch limits (~90 IDs / 4096-byte URL), pagination caps (20/page, max 60), AND-only filter logic, array-parameter serialization gotchas, and even a known isPreviewOnly bug — so an agent avoids dead ends instead of discovering them by trial.
- Discovery is the weak spot: there is no /.well-known/mcp.json, no AGENTS.md, no llms-full.txt, and the only llms.txt is the product-marketing site index rather than an API-focused map — an agent that starts from discovery files instead of the docs domain gets almost no API signal.
- The entire API sits behind a partnership agreement (credentials required), but the docs, OpenAPI spec, and Swagger UI are fully public — an agent can understand the API and generate a working client before any credentials exist.
What to change
Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.
- 01Publish an API-focused llms.txt (and llms-full.txt) at developers.epidemicsound.com that links to getting-started, auth, the guides, and spec.json — the current root llms.txt is product marketing and gives an agent no API map.
- 02Add AGENTS.md to the public demo/SDK repos (e.g. partner-content-api-demo-ios) with build, run, and auth context so coding agents have repo-level grounding.
- 03Since an official MCP server already exists, expose a /.well-known/mcp.json (or link it from a discovery file) so agents can auto-discover the MCP endpoint instead of finding it via a blog post.
- 04Add a sitemap.xml on developers.epidemicsound.com covering the docs pages — the main www sitemap omits the docs subdomain, so the API reference is not crawler-discoverable.
- 05Document idempotency behavior for write endpoints (asset upload, safelisting, usage reporting) so agents can safely retry after a timeout without duplicating side effects.
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. Three auth methods are compared in a table with step-by-step flows and a legacy migration guide. Errors return a JSON {message} body with a full status-code table mapping each code to a retry/backoff decision. Rate limits (per-second, daily app, per-user 429) and offset/limit pagination each have dedicated coverage. Idempotency keys for write endpoints (upload, safelisting, usage) are not documented.