Signals & API reference

Canonical technical documentation for integrators, researchers, and AI agents. For a compact site index see llms.txt; for the full machine-readable catalog see llms-full.txt.

REST API

Rate limits apply per client IP (default 5 requests/minute). OpenAPI schema when enabled:

MethodPathNotes
GET/healthLiveness and subsystem readiness
GET/api/v1/analyze/meUser-Agent + HTTP headers only
POST/api/v1/analyzeFull analysis; optional client_signals and ?locale=
GET/api/v1/stats/publicAnonymized benchmark aggregates (cacheable JSON)
GET/stats/feed.xmlRSS daily stats rollup
GET/stats/feed.jsonJSON Feed daily rollup

Repository API contract: see project doc/api.md on GitHub. Do not log full analysis responses in third-party systems.

Client signal probes

Collected in-browser and POSTed under client_signals (allowlisted server-side). Core tier: canvas, WebGL, audio, fonts. Extended: math, media queries, speech, WebRTC. Also: storage, automation, and protection bait probes.

Details: anti-fingerprinting guide · signal pages under /guides/signals/

Inference catalog

Each POST /api/v1/analyze response includes inferences[] with id, source (user_agent, http_headers, javascript, combined, ip_geolocation), category, title, summary, and confidence.

Full id list (no hashes or PII): llms-full.txt § Inference catalog. Deep-link to a card after analysis: /?highlight=<inference_id> (e.g. canvas_fingerprint).

Citation

Run a live analysis with full signal breakdown.

Run privacy analysis

← Protection hub