Prompt Monitoring developer & agent resources

Prompt Monitoring publishes a small, read-only information API and MCP server so agents can retrieve accurate product, pricing and guide facts. These interfaces do not expose customer data, create accounts, run monitoring checks or perform purchases.

Start here

REST examples

GET /v1/product returns product scope and measured surfaces. GET /v1/pricing returns current public pricing. GET /v1/guides lists public guides with a stable pagination shape. GET /ask?q=pricing returns a factual answer; request text/event-stream for SSE.

Errors

REST errors use application/problem+json and the RFC 9457 fields type, title, status, code, detail and instance.

Versioning and limits

Public REST paths are versioned under /v1. A successor receives at least 180 days of documented notice. Public REST (/v1/*), /ask and /mcp share 120 requests per minute per IP address. RateLimit headers report the actual remaining quota; excess requests return HTTP 429 and Retry-After. Metered responses are not cached. Static discovery documents remain cacheable without per-client quota headers. Redis provides shared counters; without Redis or during an outage, limits are best-effort per server instance. See the machine-readable guide for header semantics.

Security boundary

Public interfaces require no authentication because they only return published facts. Customer workspaces use email magic links; optional Google integrations require user-approved OAuth in the web app. Never send secrets, magic links or payment details to these public interfaces.