dreamfactorysoftware is Good to agents.
Discry independently scored how well an AI agent can discover and understand the dreamfactorysoftware 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 evaluating DreamFactory finds a genuinely rare asset: a hand-authored llms.txt (updated 2025-09-18) that opens with a canonical TL;DR, explicit 'how LLMs should use this file' instructions, a source-of-truth link map, and copy-ready REST query patterns — it can answer most questions without scraping a single page.
- The documentation itself is best-in-class for comprehension: clean Docusaurus markdown, Quick Reference tables at the top of every page (answer-first), realistic example values (emp_no 10001, 'Georgi Facello'), consistent snake_case, and end-to-end workflows (generate → secure → query → sync) rather than isolated endpoints. Comprehension scored a perfect 100.
- Discovery is what holds the grade back: no AGENTS.md in the primary repo, no llms-full.txt, and no /.well-known/mcp.json — so a coding agent cloning the repo or probing standard agent-discovery paths gets nothing beyond the README.
- The www sitemap.xml lists marketing and 'hub' pages but omits the entire docs.dreamfactory.com documentation tree (a separate subdomain), so an agent crawling the main sitemap to find API docs would miss all of them and have to rely on the llms.txt link map instead.
- DreamFactory ships an official MCP server listed on PulseMCP and Glama — an agent looking to call DreamFactory-generated APIs natively can discover a governed, RBAC-enforced tool surface, a strong signal most scanned APIs lack.
What to change
Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.
- 01Add an AGENTS.md to the primary github.com/dreamfactorysoftware/dreamfactory repo (highest-weight discovery gap at weight 3): document how a coding agent should install, configure services, and call generated endpoints. This is the single biggest available score lift.
- 02Publish a static, publicly accessible OpenAPI 3.x spec for the platform's own system APIs (db, user, system, files) — today the Swagger/OpenAPI docs only exist inside a running instance, so an agent can't retrieve a spec without standing up DreamFactory first.
- 03Add the docs.dreamfactory.com documentation URLs to a sitemap the crawler will find (either extend the www sitemap or ensure the docs subdomain's own sitemap is linked from robots.txt) so agents crawling sitemaps reach the API reference.
- 04Serve /.well-known/mcp.json advertising the official DreamFactory MCP server and its tool declarations, turning an already-published capability into a machine-discoverable one at the standard path.
- 05Generate an llms-full.txt that inlines the core database-API, auth, and querying docs as markdown, so agents get a single comprehensive payload without following links to individual Docusaurus 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.
Execution characteristics are exceptionally well documented for a generated-API platform. Errors return machine-parseable JSON ({"error": {code, context, message}}); auth spans API key + JWT plus five SSO/identity protocols with identity passthrough; pagination is limit/offset with include_count. Rate limiting is documented as a commercial-tier feature (configurable) but specific default limits and header names aren't published. Idempotency keys are not mentioned, though PUT (full replace) vs PATCH (partial) semantics are clearly distinguished.