Metaplane is Good to agents.
Discry independently scored how well an AI agent can discover and understand the Metaplane 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 · 91/100What we found
- An agent that lands on the marketing domain (metaplane.dev) hits a dead end — no llms.txt, no docs in the sitemap — but the real docs at docs.metaplane.dev ship a well-structured, API-focused llms.txt that indexes every guide and endpoint in markdown, so a persistent agent is rewarded richly.
- Comprehension is excellent (91/A): every endpoint carries a task-oriented description ('Create a new monitor on a given absolute path...'), docs are natively markdown via .md URLs, and the full OpenAPI 3.0.1 definition is embedded on each reference page with detailed schemas, enums, and default values.
- The OpenAPI spec is public and versioned but fragmented — it is exposed per-endpoint inside reference pages rather than as a single downloadable openapi.json (the standard path 404s), forcing an agent to stitch it together instead of ingesting one canonical file.
- Discovery is dragged down by closed-source SaaS gaps an agent can't work around: no AGENTS.md (no public repo), no .well-known/mcp.json, no llms-full.txt, and no MCP-registry listing — so an agent discovering Metaplane via a registry or GitHub would never find it.
- An agent can grasp core capabilities fast — the llms.txt index (~13.5KB) plus a 3.3KB getting-started page convey the whole product in well under 5,000 tokens, and rich Monitors-as-Code YAML examples use realistic column names, cron schedules, and SQL.
What to change
Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.
- 01Publish a single canonical OpenAPI file at a standard path (e.g. docs.metaplane.dev/openapi.json) and link it from llms.txt so agents can ingest the whole spec in one fetch instead of scraping per-endpoint fragments.
- 02Add an llms.txt (and a redirect or copy of the docs llms.txt) on the primary metaplane.dev domain — agents that start at the brand domain currently find no machine-readable entry point to the API.
- 03List the Metaplane API/MCP in at least one major registry (Glama, Smithery, or PulseMCP) so agents performing tool discovery can find it; publish a lightweight MCP wrapper or a .well-known/mcp.json to declare available tools.
- 04Add a dedicated Authentication + Errors + Rate Limits reference page with explicit recovery steps for the top failures (401/invalid key, 429, 400 validation) — e.g. 'on 429, back off using the Retry-After header' — turning today's partial error guidance into agent-usable self-correction.
- 05Generate a docs sitemap.xml on docs.metaplane.dev (or add the docs URLs to the marketing sitemap) so crawlers that rely on sitemaps can enumerate the reference and guide pages.
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.
The public API is an OpenAPI 3.0.1 spec secured by an Authorization-header API key. Error responses carry a structured `errorMessage` field and several endpoints document 404 conditions inline, but there is no dedicated error-format or rate-limit reference. Pagination is documented for history/tag endpoints (createdAt-based, 500 rows/page); idempotency keys are not mentioned.