LaunchDarkly is Agent-Ready to agents.
Discry independently scored how well an AI agent can discover and understand the LaunchDarkly 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 · 88/100Comprehension
55% of score · 96/100What we found
- An agent can consume every documentation page as clean Markdown simply by appending .md to the URL — and each page's header advertises this, the llms.txt index, and a docs MCP server endpoint. This is best-in-class agent affordance: the docs were clearly built for machine readers, not just humans.
- LaunchDarkly is one of the most agent-forward devtools scanned: an official hosted+local MCP server (listed on PulseMCP and Glama), an AGENTS.md in the ldcli repo, a public agent-skills/playbooks repo, a GitHub Copilot flag-cleanup agent, and native 'AI Configs' / AgentControl API surfaces.
- The REST API is fully discoverable and machine-usable: a 2.8 MB OpenAPI 3.0.3 spec is publicly served at app.launchdarkly.com/api/v2/openapi.json covering 385 operations across 243 paths, 100% with summaries and 99% with descriptions.
- robots.txt allows all crawlers (only an internal /api/fern-docs/ path is disallowed) and the sitemap lists 1,396 doc URLs — an agent searching for LaunchDarkly will find it and is never blocked.
- The biggest discovery gap is llms.txt sizing: the file is a 191 KB site-wide link dump (well above the ~50 KB agent-friendly target) and llms-full.txt is byte-identical to it rather than containing inlined full content — so an agent pulling the index pays a heavy token cost for capability discovery.
What to change
Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.
- 01Trim llms.txt to an API/capability-focused index under ~50 KB (group by product area with short section descriptions) and reserve the exhaustive link list for llms-full.txt — currently the two files are identical, wasting the convention.
- 02Make llms-full.txt actually 'full': inline the core Markdown content (overview, auth, key endpoints, error/rate-limit behavior) rather than duplicating the link index, so an agent can ingest usage patterns in one fetch.
- 03Publish a /.well-known/mcp.json pointing at the existing docs MCP server (/docs/_mcp/server) so MCP-aware clients can auto-discover the integration without reading prose — the server already exists; only the discovery manifest is missing.
- 04Document idempotency support (idempotency keys for POST/PUT) in the OpenAPI description, or explicitly state that retries are safe via the existing versioning model — agents performing automated writes need this to retry safely after 429/5xx.
- 05Add a concise (<5,000 token) 'API overview' / capability-map page (the /docs/api/overview and rest-api intro paths currently 404 as .md) so an agent can grasp core capabilities without parsing the 191 KB index or 50 KB getting-started page.
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.
Exceptionally complete execution documentation in the OpenAPI spec description: API-key auth via the Authorization header (personal/service tokens, TLS 1.2 minimum), JSON error responses with a documented error-code table, rate limiting with named headers (X-Ratelimit-Global-Remaining, X-Ratelimit-Reset) plus Retry-After and explicit jitter/backoff guidance, HATEOAS pagination via _links (self/next/first/last) with summary-vs-detailed representations, and dated API versioning with clearly marked beta resources. Idempotency keys are not documented.