PowerAPI is Poor to agents.

What we found

• —An agent that reaches the docs finds surprisingly solid content — a real endpoint reference (Agents API with parameter tables, curl + JSON examples), an error-codes page with explicit recovery actions (429 → Retry-After, 500 → exponential backoff), and rate-limit headers — but almost none of it is URL-addressable: every section (Authentication, Agents API, Error Codes, Rate Limits, Webhooks) lives behind client-side tab clicks on a single /docs SPA, so a standard scrape sees only the Quickstart.
• —The developer trust chain is broken: the 'View on GitHub' and 'open source' links point to github.com/powerapi, an archived, unrelated org (PowerSchool client libraries, archived Aug 2023), and the /developers page links to /docs/api and /docs/sdks which both 404. An agent following the site's own links to verify the 'open-source' claim hits dead ends everywhere.
• —llms.txt exists and is clean, well-structured markdown with per-link descriptions — a genuine positive — but it is product-page-focused and points agents at /docs and /developers rather than at any API-specific content, and there is no llms-full.txt to compensate for the un-scrapeable SPA docs.
• —No machine-readable API surface exists at all: no OpenAPI/Swagger spec at any standard path or linked from docs, no .well-known/mcp.json, no MCP registry listings, and no AGENTS.md — an agent must reconstruct the entire API from prose it can only reach by simulating button clicks.
• —Where the docs are readable, they are agent-friendly: task-oriented endpoint descriptions, realistic Stripe-style example values (agt_abc123, pk_live_ keys, EUR pricing config), two complete multi-step workflow walkthroughs (deploy → execute → poll → consume output), and answer-first endpoint layout.

What to change

Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.

1. 01Make every docs section URL-addressable (e.g. /docs/authentication, /docs/agents-api, /docs/errors) or publish an llms-full.txt containing the full docs as markdown — this single change would take most of the documentation from invisible to fully agent-consumable.
2. 02Fix or remove the broken developer-trust links: point 'View on GitHub' at a real owned org, and fix the 404ing /docs/api and /docs/sdks links on the /developers page (they undermine every other claim the site makes).
3. 03Publish an OpenAPI 3.x spec at powerapi.com/openapi.json covering the documented v2 Agents/Skills/Locations/Integrations endpoints — the endpoint reference content already exists, it just isn't machine-readable.
4. 04Sharpen llms.txt into an API-focused index: link directly to the quickstart, authentication, endpoint reference, and error-codes content (once URL-addressable) instead of only marketing pages.
5. 05Standardize field casing in API request/response bodies — snake_case (created_at, per_page) and camelCase (checkInterval) are mixed within the same JSON payloads, which invites malformed agent-generated requests.