PredictLeads is Needs Work to agents.
Discry independently scored how well an AI agent can discover and understand the PredictLeads 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 · 71/100Comprehension
55% of score · 72/100What we found
- An agent hitting predictleads.com/llms.txt gets one of the strongest llms.txt files in the corpus: API-focused, structured, and it links directly to the OpenAPI spec, MCP server endpoint, auth setup, rate limits, and pricing — nearly everything needed to start integrating.
- PredictLeads ships an official hosted MCP server (mcp.predictleads.com) published in the official MCP registry (com.predictleads), so MCP-aware agents can query the data natively without writing REST glue.
- A public OpenAPI 3.1 spec at docs.predictleads.com/schemas/open_api_schema.json gives agents a machine-readable contract for all discovery and company-level endpoints.
- The reference docs are a JS-rendered single-page app: all 613 request/response code blocks come back empty when converted to markdown, so an agent scraping the docs sees Shell/Python example tabs but no runnable example content.
- Error documentation is one-line status descriptions ('402 Exceeded the monthly request limit', '429 Too many requests') with no recovery steps in the reference — the only retry guidance (Retry-After on 429) lives in llms.txt.
What to change
Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.
- 01Render code examples server-side (or publish an llms-full.txt with inline request/response examples) so the 613 currently-empty example blocks survive HTML-to-markdown conversion for agents.
- 02Add actionable recovery guidance to error docs: document Retry-After handling for 429, what to do on 402 quota exhaustion, and the difference between 403 rate-over-limit and auth failures.
- 03Add an AGENTS.md to the PredictLeads GitHub org repos with integration context for coding agents.
- 04Document the numeric rate limit (60 req/s) in the reference docs themselves, not only in llms.txt, and add a known-limitations section covering what the API cannot do.
- 05Fix docs discoverability in sitemaps: docs.predictleads.com/sitemap.xml returns 406 and the main-domain sitemap only lists marketing 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.
JSON:API response format with error schemas defined in the public OpenAPI 3.1 spec; rate limit (60 req/s, 429 + Retry-After) and monthly credit quota (402 on exhaustion) documented; page-based pagination with limit default 100 / max 1000. Idempotency is not mentioned, and error examples in the reference render as empty code blocks when scraped.