Integrate FlareWatch's Flare & Songbird analytics — validator scores, FTSO data-provider metrics, network stats, and prices — into your app or AI agent. The REST API is public and keyless (rate-limited); for tool-native access, point any MCP client at the MCP server. No signup, no API key, no "contact sales."
Every endpoint is a plain GET returning JSON. No auth header needed.
# Live FLR / SGB prices (FTSO-primary) curl https://www.flarewatch.io/api/cache/prices # All Flare validators (NodeID -> name) curl https://www.flarewatch.io/api/validators/name-map # One validator's full score + per-dimension breakdown curl https://www.flarewatch.io/api/validators/NodeID-6aKkPwTjFa3A5SH9H678r48MmNwyKFjQj/score-breakdown
GET /api/statsPlatform coverage counters.GET /api/cache/pricesLatest FLR & SGB prices (multi-currency + market cap).GET /api/cache/ftso-providersScored FTSO data providers on Flare.GET /api/cache/sgb-ftso-providersScored FTSO data providers on Songbird.GET /api/validators/{nodeID}/score-breakdownFull FlareWatch score + per-dimension breakdown for one validator.GET /api/validators/name-mapNodeID -> display-name map for all validators.Full machine-readable spec (typed schemas, operationIds, examples): /openapi.json (OpenAPI 3.1). LLM-oriented index: /llms.txt.
The Model Context Protocol server exposes the same data as callable tools over Streamable HTTP, so Claude, ChatGPT, Cursor, and any MCP-compatible client can query FlareWatch natively. Public read tools need no auth; wallet-scoped tools use OAuth 2.1 + PKCE with one-time consent.
# Endpoint https://mcp.flarewatch.io/api/mcp (MCP Streamable HTTP, spec 2025-06-18) # npm bridge npx @flarewatch/mcp
Tool list, security model, and a live "Try it" sandbox: the MCP page. Discovery manifest: /.well-known/mcp.
The REST API is read-only and keyless— no token, no signup. It is rate-limited per IP (see below). Only the MCP server's wallet-scoped tools require authentication, via OAuth 2.1 + PKCE (discovery at /.well-known/oauth-authorization-server); public MCP tools need none.
Every response carries standard rate-limit headers so you can self-throttle in real time:
RateLimit-Limit: <requests allowed in the window> RateLimit-Remaining: <requests left> RateLimit-Reset: <seconds until the window resets> # On a 429, additionally: Retry-After: <seconds to wait>
For heavy or bursty workloads, prefer the MCP server, which is built for tool-call volume.
Errors are always JSON — never an HTML page — with a stable, machine-readable code and a recovery hint:
{
"error": "Validator NodeID-... not found in current cache.",
"code": "VALIDATOR_NOT_FOUND",
"hint": "Confirm the NodeID is an active Flare P-Chain validator; list all via /api/validators/name-map."
}Codes include RATE_LIMITED, INVALID_NODE_ID, VALIDATOR_NOT_FOUND, CACHE_UNAVAILABLE, and INTERNAL_ERROR.
This is API v1. The unversioned paths above serve v1 and are stable; every response stamps X-API-Version: 1. A breaking change ships under a new major version, never in place. Deprecations are announced at least 90 days ahead and signaled on responses with the standard Deprecation and Sunset headers (RFC 8594) pointing back here — so an agent integrating today will never be broken without warning.
Questions or a deprecation you need lead time on? Contact us.