Opencodeopencode.ai
Opencode's API program has a standout strength in agent understanding and API contract quality, giving AI agents a reliable surface to build against. However, two areas need urgent attention: developer experience is undermined by sparse descriptions across the vast majority of endpoints and parameters, and agent usability is critically weakened by the complete absence of idempotency guidance and rate-limit signaling — meaning agents retrying mutating operations risk silent data duplication, and have no signal before they get throttled.
API DesignA clean, typed, well-governed API contract agents can reason about4 pass1 warn0 fail94A+
| Signal | Points | Findings | Rationale | |
|---|---|---|---|---|
| pass | Schema coverage & depthvia sdk | 25/25 | 3/3 enums are named types (not bare strings). Investigated: sdk 100%, spec 50%. | Typed, complete request/response schemas are what make agent function-calling possible. |
| pass | Auth declared & discoverablevia spec | 25/25 | This API requires no credentials and says so — the spec declares `security: []`, meaning no credentials are required (spec). An agent can call it directly. Investigated: spec 100%, docs 100%, mcp 100%, cli 0%. | Agents can only call an API when they can determine its auth posture — a declared scheme, or an explicit statement that none is required. |
| warn | Machine-readable, versioned contractvia spec | 18.8/25 | The API sets a version ("1.0.0") but exposes no versioning scheme in the URL, a header, or the media type, so an agent can't pin to a specific version. Investigated: spec 75%, docs 50%, cli 0%. Fix: Pick a versioning scheme (URL path /v1/, Accept: application/vnd.x+json, or a version header) and apply it consistently. | A current OpenAPI version with a declared versioning scheme lets agents reason about the contract. |
| pass | Security & governance hygienevia spec | 15/15 | No credential-shaped strings detected in spec. Investigated: spec 100%, sdk 83%, wellknown 0%. | No leaked secrets, no critical lint violations, no OWASP API Top-10 spec smells, and a published vulnerability-disclosure channel. |
| pass | Example coveragevia docs | 10/10 | Docs show code samples in 4 language(s) (top: shell, javascript, typescript), and 10 of 20 sampled pages offer ≥ 2 languages side by side. Investigated: docs 100%, sdk 100%, spec 0%, cli 0%. | Examples carry shape semantics schemas under-specify — for humans and agents alike. |
Developer ExperienceThe context both developers and agents need to integrate fast — onboarding, code samples, complete descriptions and worked examples3 pass0 warn2 fail68C
| Signal | Points | Findings | Rationale | |
|---|---|---|---|---|
| pass | Self-service developer portalvia docs | 29/29 | The API requires no credentials (the spec declares `security: []`, meaning no credentials are required (spec)), so there is nothing to sign up for and no sales gate to clear. Investigated: docs 100%. | A first call without a human in the loop — self-serve credentials, a free tier, or an API that needs none. |
| pass | Code samples in docsvia docs | 18/18 | Docs show code samples in 4 language(s) (top: shell, javascript, typescript), and 10 of 20 sampled pages offer ≥ 2 languages side by side. Investigated: docs 100%. | Multi-language samples shorten time-to-first-call. |
| pass | Changelog publishedvia spec | 13/13 | changelog is documented on the docs site at https://opencode.ai/changelog, even though it isn't declared in the API spec. Investigated: spec 100%, docs 100%. | A published changelog lets partners track changes without surprise. |
| fail | Quickstart presentvia docs | 0/25 | No quickstart or getting-started page found at the conventional docs paths, so new developers and agents have no obvious first step. Investigated: docs 0%. Fix: Publish a single-page quickstart that goes from zero to a first successful API call. | A quickstart is the fastest path from landing page to first successful call. |
| fail | Description completenessvia spec | 0/15 | Only 33% of operations and parameters have descriptions (124 of 187 ops, 0 of 414 params), leaving an agent to guess what most endpoints do. Investigated: spec 0%. Fix: Descriptions are the primary semantic signal agents use for tool selection. Each operation should explain what it does and reference its parameters; each parameter needs a real description. | Complete descriptions are the context humans and agents need to use endpoints. |
Agent DiscoveryPartners and their agents can find your APIs — llms.txt, registries, crawlable and reachable docs2 pass1 warn1 fail73B
| Signal | Points | Findings | Rationale | |
|---|---|---|---|---|
| pass | Docs reachable, not hard auth-gatedvia docs | 30/30 | All 1 pages are publicly accessible. Investigated: docs 100%. | Agents can only index and fetch docs they can reach — past auth gates and over correct HTTP semantics. |
| pass | Registry & SDK presencevia docs | 28/28 | Indexed on Context7 (anomalyco/opencode, 11195 snippets). Investigated: docs 100%, cli 100%, sdk 50%, mcp 0%, wellknown 0%. | Listing in MCP registries and publishing SDKs puts the API where agents and their tooling look. |
| warn | Crawlable / AEOvia wellknown | 9/12 | Sitemap has 706 entries but no lastmod values, so crawlers can't tell what changed recently. Investigated: wellknown 75%. Fix: Add accurate <lastmod> values to sitemap entries — AI-powered search uses them to prioritize crawling. | Bots allowed plus a fresh sitemap make docs findable by agent crawlers. |
| fail | llms.txt present, valid & comprehensivevia docs | 0/30 | No llms.txt found at any candidate location (https://opencode.ai/docs/llms.txt, https://opencode.ai/llms.txt); 1 failed to fetch. Investigated: docs 0%, sdk 0%. | A valid, comprehensive llms.txt is the machine-readable entry point for agents. |
Agent UnderstandingAgents can correctly interpret your APIs — machine-readable errors, consistent descriptions, structured data, parseable docs3 pass2 warn0 fail92A+
| Signal | Points | Findings | Rationale | |
|---|---|---|---|---|
| pass | Machine-readable errors (RFC 9457)via docs | 28/28 | Documents 7 distinct 4xx/5xx error response codes, so an agent can branch on specific failures instead of guessing. Investigated: docs 100%, spec 50%, cli 0%. | RFC 9457 problem details and a documented error-code inventory let agents parse failures without burning tokens. |
| pass | Operation purpose clarityvia spec | 25/25 | 94% of operations (175 of 187) have both a clear summary and a descriptive name, so an agent can pick the right endpoint. Investigated: spec 100%. | Agents select the right endpoint from its summary + operationId; clear, named operations make tool-selection reliable — the strongest driver of correct tool choice. |
| pass | Agent-navigable, token-efficient docsvia docs | 22/22 | All 1 pages contain server-rendered content. Investigated: docs 100%. | Server-rendered, clean, small-footprint docs are what an agent can cheaply fetch and parse correctly. |
| warn | Agent instructions file (AGENTS.md)via sdk | 5/10 | No agents.md or skill.md context file found (nice-to-have for agent operation). Investigated: sdk 50%, cli 0%, wellknown 0%. Fix: Add an agents.md or skill.md describing how an agent should use this SDK (auth setup, key operations, gotchas). | An AGENTS.md gives coding agents explicit setup, auth, and usage instructions to interpret and operate the API — beyond llms.txt's link index. |
| warn | Docs structured datavia docs | 4/8 | Only OpenGraph/meta tags across 3 assessed pages — no JSON-LD for AI answer engines to cite. Investigated: docs 50%. Fix: Emit JSON-LD with Article/TechArticle @type and dateModified in the rendered docs head. | Structured data (JSON-LD/schema.org) on docs pages gives agents an unambiguous parse target and is what answer engines cite. Detected on the JS-rendered head (Firecrawl) for a bounded page budget, so JS-injected JSON-LD is now caught; pages we can't render are excluded rather than failed. |
| na | Description consistency across surfaces | —/7 | Only 1 surface description(s) with ≥6 tokens available; need at least 2 to compare. | Every surface tells the same story about what the product is. |
Agent UsabilityAgents have the context to use your APIs reliably, not just find them2 pass1 warn2 fail56D
| Signal | Points | Findings | Rationale | |
|---|---|---|---|---|
| pass | Pagination documented & consistentvia spec | 22/22 | Pagination parameters are present on the 50 list endpoint(s), so an agent can page through results. Investigated: spec 100%, docs 100%. | Consistent, documented pagination lets agents traverse collections. |
| warn | Sandbox separationvia sdk | 10/20 | No test-mode flag or sandbox environment is exposed. Investigated: sdk 50%, spec 0%, docs 0%. Fix: Expose a test-mode flag or sandbox environment so agents can exercise the SDK safely. | An isolated environment lets agents exercise destructive operations safely. |
| pass | Runnable collection with test scriptsvia sdk | 9/9 | README is present and includes a runnable quickstart. Investigated: sdk 100%, platform 0%. | A public, maintained collection with assertions is runnable truth agents validate against. |
| fail | Idempotency documentedvia spec | 0/27 | Only 0% of mutating operations (0 of 100) document idempotency, so an agent's retries can create duplicate writes. Investigated: spec 0%, docs 0%, sdk 0%, cli 0%. Fix: Agents retry on failure; without idempotency, retries create duplicates. Document Idempotency-Key on every POST/PUT/PATCH/DELETE. | Documented idempotency lets agents retry safely. |
| fail | Rate-limit signalingvia spec | 0/22 | No rate-limit response headers are documented, so an agent can't tell when it is approaching a limit and will get throttled. Investigated: spec 0%, docs 0%. Fix: Add x-ratelimit-* response headers and document retry-after semantics. | Machine-readable rate-limit headers let agents throttle adaptively. |
Resources Discovered
The public resources we found for Opencode — the evidence behind the score. All discovered from public sources; nothing here requires access to your systems.
| Agent hints | Context7 (11,195) |
|---|---|
| APIs analyzed | 1 — opencode |