Swell is Poor to agents.
Discry independently scored how well an AI agent can discover and understand the Swell 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 · 29/100Comprehension
55% of score · 63/100What we found
- Swell has zero agent-discovery infrastructure despite having a well-structured developer docs site — no llms.txt, no OpenAPI spec, no AGENTS.md means agents cannot find or understand this API without manual research
- The docs content quality is solid when found — task-oriented descriptions, consistent naming, clean hierarchy — but an agent would never discover it through standard channels
- A community MCP server exists (devkindhq/swell-mcp by official Swell partners) indicating demand for agent integration that the official docs don't yet support
- No error documentation or rate limit information is publicly visible, leaving agents with no self-correction capabilities
- The quickstart guide is conceptual rather than actionable — it explains what Swell can do but doesn't walk through API calls step by step
What to change
Prioritized by impact on discoverability. You (or your docs platform) deploy these — Discry never touches your API.
- 01Add llms.txt covering the Backend API and Frontend API with structured descriptions of core resources (products, orders, customers, subscriptions)
- 02Publish an OpenAPI spec for the REST API — this is the highest-impact single change (weight 5) for agent discoverability
- 03Add comprehensive error documentation with error codes, response format, and recovery guidance for common scenarios
- 04Document rate limits and API boundaries so agents can implement throttling
- 05Create end-to-end workflow guides (e.g., 'Create a product with variants and pricing' or 'Process an order from cart to fulfillment')
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.
Store ID + secret key authentication via official Node and PHP libraries. Custom wire protocol on port 8443 for performance. No error format documentation found. No rate limits documented. Pagination supported but details are within endpoint documentation. No idempotency mentioned.