Tripl

Keep your product analytics honest.

Docs · Quick start · Try the demo

_{A volume spike on one event, attributed to the slice that caused it.}

---

tripl keeps your tracking plan — the events, fields, and values your product is supposed to send — and checks it continuously against the events that actually land in your warehouse. When the two diverge, it tells you.

• Plan as code. One catalog for every event, field, and value; changes are made on branches, reviewed, and merged.
• Checked against real data. Scans read your warehouse tables and propose events and fields; reconciliation flags what is undocumented and what stopped arriving.
• Anomalies with a cause. Seasonal baselines per event, schema and value drift, release regressions — each signal broken down by the slice that moved.
• Alerts where your team works. Slack, Telegram, email, webhooks, Jira, Linear.

No SDK. tripl reads from the warehouse you already have — ClickHouse, BigQuery, or PostgreSQL — and never writes to it.

Built for product managers, analysts, and data engineers who own a tracking plan.

---

Quick start

cp .env.example .env
docker compose -f compose.dev.yaml up --build

Open http://localhost:5173, create the first account, and click Generate demo project. No warehouse is needed: the demo builds a project with events, fields, a week of metrics, and a few anomalies, backed by a local synthetic warehouse. Scans, metric collection, anomaly detection, and reconciliation run against it for real (what is synthetic).

To use your own data, add a warehouse under Settings → Data sources and follow the Quick Start guide.

Local dev stack	URL
App	http://localhost:5173
API	http://localhost:8000
API reference (interactive)	http://localhost:8000/docs

---

What's inside

tripl is organised around three jobs: Plan, Observe, and Govern.

📐 Plan Every event, field, and value in one searchable catalog. Changes go on a branch and are reviewed before they merge, like a pull request for your tracking plan.

📊 Observe Anomaly detection that learns each event's daily and weekly rhythm, plus schema drift, release regressions, and a metrics catalog. Every signal shows which slice of the data moved.

🛡️ Govern Reconciliation answers two questions: what's documented but no longer arriving, and what's arriving but was never documented. Coverage, an audit log, and roles round it out.

Automation. Scoped, revocable API keys; a CLI (pip install tripl); and an MCP server (tripl-mcp) through which an LLM agent can search the plan and propose changes on a branch.

See Concepts for how the pieces fit together.

---

Deployment

The default compose.yaml runs the published release image: set your secrets, run docker compose up -d, and the app comes up on :8000. The deployment guide covers the rest, and cutting a release is one command, bin/release.sh (see the release process).

---

Documentation

📖 The full documentation lives at vladenisov.github.io/tripl (sources under website/docs/). Good places to start:

• Quick Start — from docker compose up to a scanned plan, your own metrics, and a first alert.
• Concepts — tracking plans, events, branches, and monitors, in plain language.
• User guide — a hands-on walkthrough from your first project to a working alert.
• Variables & templates — documented values, source bindings, per-event overrides, and value drift.
• Agent & API guide — letting an LLM agent or a script read and update the plan.

---

Contributing

tripl is a FastAPI + PostgreSQL backend, a Celery worker that talks to your warehouses, and a React frontend, all runnable locally with Docker Compose.

• CONTRIBUTING.md — local setup and commands. The root Makefile collects the common ones; run make to list them.
• Architecture — how the system is built, and why.
• AGENTS.md — a navigation map of the repo for coding agents.