Every integration is the same two-value change — point the client at Spro's base URL and use your Spro API key. Requests, responses, streaming, and error formats stay exactly what your tool already expects.
Pick by what you're building:
OpenAI SDKChat Completions, Responses API, embeddings, images — full OpenAI compatibilityAnthropic SDKThe Messages API is Anthropic-compatible — keep your Claude code unchangedCoding AssistantsCursor, Claude Code, Cline, Aider on every model in the catalogOpenAI Agents SDKBuild agents — with traces flowing into your Spro dashboardRaw HTTPNo SDK — curl, httpx, fetch, any HTTP client in any languageDocs MCP ServerLet your AI agent search and read these docs over MCP
The universal pattern
Whatever you use, the setup is identical:
Base URL: https://spro.hridaai.com/v1API key: spro-xxxxxxxx (starts with spro-)Model: <provider>/<model> (e.g. anthropic/claude-sonnet-4) or a bare nameModel naming is the only concept to learn — see Models.
What you get on every surface
GuardrailsPII, prompt injection, jailbreaks — inspect every request on every endpoint.Usage & Cost TrackingToken counts, cost, and latency per project and per API key.TracingTrace IDs on every request, resolved-model headers, and agent spans.BYOKBring your own provider keys — your pricing, your quotas.Automatic FallbacksIf a provider fails, Spro routes to the next best option.
