Trace is an open-source, self-hostable service that recognises returning devices and flags risky ones. A page loads the browser SDK, which collects device signals and sends them to your server. The server answers with a stable visitorId, a suspect score and the reasons behind it. Your backend then verifies the event with a secret key before trusting it.
It runs entirely on your infrastructure. No data leaves your deployment unless you configure an external provider.
What you get
- Visitor identification that survives private windows, cleared cookies and minor browser updates, with a confidence value.
- Signals: bot and automation, browser tampering, Tor, datacenter IPs, IP blocklists, velocity, incognito, developer tools, virtual machines, privacy-focused browsers and high-activity devices. Every event explains which signals contributed to its score.
- Server-side verification (
GET /v1/events/:eventId), visitor history, deletion for privacy requests, signed webhooks. - A dashboard for projects, keys, events, visitors, usage and webhooks.
- SDKs: browser (npm or one script tag), React, Node.
How it fits together
Browser (SDK) --POST /v1/identify, public key--> server --> Postgres, Redis
Your backend --GET /v1/events/:id, secret key--> server
Dashboard --session cookie--> serverHonest expectations
Identification accuracy is measured on labeled and synthetic data and published on the accuracy page. It is well below commercial services on real traffic, and client-side signals can be spoofed by a determined attacker. Use the score as one input next to your own defences, and start in shadow mode.
Next: the 15-minute quickstart.