People who run their own infrastructure

Developers

One process, embedded database, zero runtime dependencies. Apache-2.0 core — read it, change it, run it on your own box. And an MCP server so your coding agent can query it directly.

Get started free Read the docs

Questions it answers

Is it really a one-command install?

Yes. With Bun installed, `bun run cloud` is enough — no second service, the database is embedded. No three-service Plausible setup, no Kafka+ClickHouse stack.

Can my AI agent query the analytics?

Yes — there is an MCP server. It is read-only on purpose, and every metric comes back with the SQL, the parameters and the raw rows, so the agent can cite evidence instead of asserting a number.

Can I get at the data directly?

Self-hosted, the database file is yours — query it with SQL. In the cloud, /api/stats returns the evidence bundle as JSON, including each metric's query and raw result.

How big is the tracker?

2.4 KB gzipped, including Core Web Vitals and error capture. No cookies, no stored identifier, no third-party requests. Do Not Track is respected by default.

I don't want data going to an LLM.

No model is enabled by default; the digest is generated deterministically. If you want one, point it at Ollama on your own machine — nothing leaves the box.

Setting it up

Run it yourself

git clone <repo> /opt/vitrus && cd /opt/vitrus
bun install
bun run build:tracker
bun run cloud

Connect your agent over MCP

// claude_desktop_config.json / .mcp.json
{
  "vitrus": { "url": "https://app.vitrus.dev/mcp" }
}

Read-only. Every answer includes the query that produced it.

Pull the evidence bundle

curl "https://app.vitrus.dev/api/stats?site=SITE_ID&days=7" \
  -H "cookie: vitrus_session=..."