Datanomik is Needs Work to agents.
Discry independently scored how well an AI agent can discover and understand the Datanomik 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 · 67/100Comprehension
55% of score · 69/100What we found
- An agent lands well: the docs home page and every page carry an explicit 'For AI agents' pointer to https://docs.datanomik.com/llms.txt, which is a clean, API-focused index linking each guide and endpoint to a Markdown (.md) version.
- Every reference page embeds a complete, valid OpenAPI 3.1 definition (servers, security scheme, described query params, enums like ISO country codes), so an agent can machine-read the full request contract without a separate spec download.
- Response examples are genuinely realistic — real-looking UUIDs, CNPJs, BRL amounts, ISO dates and paginated envelopes — which lets an agent predict payload shapes accurately.
- The biggest gap is error handling: there is no error-code catalogue and no recovery guidance anywhere, so an agent that hits a 4xx/5xx gets no documented way to self-correct.
- Discovery is weakened by fragmentation across two API surfaces (api.datanomik.com openbanking-api and treasury-api.datanomik.com remuneration-api) with duplicate-named endpoints, no AGENTS.md, no MCP registry listing, and no docs sitemap.
What to change
Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.
- 01Publish an errors reference (HTTP status codes, error response schema, and concrete recovery steps — e.g. what to do on 401 auth failure, 429, or an invalid linkId) so agents can self-correct rather than fail blindly.
- 02Expose a single consolidated, downloadable OpenAPI spec (one file covering both the openbanking and remuneration/treasury APIs) and link it from llms.txt, so agents can load the whole contract in one fetch.
- 03Add 2-3 end-to-end workflow guides beyond the 3-step quickstart (e.g. connect institution -> obtain link_id -> list accounts -> pull transactions, and a payments/PIX flow) with error handling, not just isolated endpoint pages.
- 04Add descriptions to endpoints that currently have empty or name-echoing summaries (e.g. List Transactions, List Investments) and de-duplicate the repeated 'List Accounts'/'Detail Account' entries so agents aren't confused about which surface to call.
- 05Publish a docs-domain sitemap.xml that includes the API reference pages, and add an AGENTS.md to any public SDK/quickstart repo to strengthen discovery.
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.
Auth is documented as HTTP Basic using a secretId/secretPassword pair generated in the dashboard, and pagination is well-covered (page/size params with a dedicated page). However, error formats, error codes, rate limits, and idempotency are undocumented — an agent would only discover failure behavior by hitting it live.