Pug - Open Source Product Analytics and User Profiles

Open beta: free

Pug records what happens in your product: page views, clicks, scrolls, custom events. Every one is tied to a person.

Ask questions with trends, funnels, retention, and user flows.

Self-host the whole stack, or use the free cloud.

GDPR EU‑hosted Cookieless by default

user-09996

ext cust_009996

Created May 30 First seen 123d ago Last seen 4d ago Android 11 New York City, US

Identified properties1 trait

nameDiana Jones

Recent activitylast 10 events

app_close 4d ago
scroll 4d ago 72%
page_view 4d ago
app_open 4d ago
checkout_started 4d ago USD 416.21
add_to_cart 4d ago prod-0198
search 4d ago shirt

Get started

First events in two minutes

Initialize the SDK and auto-tracking starts immediately. No event plan, no instrumentation sprint. It starts cookieless: events flow, nothing is written to the device. Grant consent and identify() merges a signed-in person's anonymous history into their profile.

import { init, identify, track } from '@pug-sh/browser'

init('your-project-id', { apiKey: 'pub_your_key' })

// Page views, clicks, scrolls, form interactions, rage clicks,
// and dead clicks are captured automatically from here on.
identify('user_123', { email: 'ada@example.com', plan: 'pro' })
track('signup', { plan: 'pro' })
track('order_completed', { amount: 49.0, currency: 'USD' })

import 'package:pug_flutter/pug_flutter.dart';

await Pug.init('your-project-id', const PugOptions(apiKey: 'pub_your_key'));

// Auto-tracking starts immediately. Identify users on sign-in:
await Pug.identify('user_123', traits: {'plan': 'pro'});

// Typed methods for well-known events (custom props go in extras):
Pug.track.signup(extras: {'plan': 'pro'});

import { Pug } from '@pug-sh/node'

const pug = new Pug({ apiKey: process.env.PUG_API_KEY }) // private key, prv_…

// Server events name the user explicitly - no ambient session.
pug.track('user_123', 'order_completed', { amount: 49.0, currency: 'USD' })
await pug.identify('user_123', { email: 'ada@example.com', plan: 'pro' })
// Private keys can also read: profiles and insight queries.
const profile = await pug.profiles.getByExternalId('user_123')

Captured automatically

page_view every route change
click element, text, selector
scroll max depth per page
form_submit form interactions
rage_click frustration signal
dead_click clicks that did nothing

Plus your own

signup name any action you care about
order_completed with properties, amount, currency
video_play queryable like any built-in event

Added on ingest

Geo (country, region, city)
Device & browser
UTM parameters

Every SDK Framework guides What autocapture covers

Live

See events arrive in real time

Every event shows up seconds after it happens, on a live map and in a stream, with the page, device, and location already attached.

Ship a tracking change and know within seconds that it fires
Watch a launch land as it happens: how many people, from where, on which pages
Filter to the event behind a new feature and watch the first real users hit it

Everything you can ask

Unified profiles

One profile per person, however they arrive

The profile you saw at the top of this page is the core of Pug: every session, trait, and event a person ever generated, in one searchable timeline.

Before sign-in, events accrue to an anonymous ID
identify(userId) merges that history into one profile
Works across devices, web today, app tomorrow: same person
Traits (plan, email, anything) live on the profile and filter every insight

Analytics per person How identity resolution works

Unified profile user_123

page_view anon
scroll anon
add_to_cart anon
identify('user_123') history joins
signup
order_completed

Insights

Six ways to turn events into answers

Every insight runs over raw events in ClickHouse. Filter by any property, break down by any dimension, and get answers in seconds.

Trends

Time-series of any event. Count, unique users, avg per user, or sum/avg/min/max of a numeric property. Break down by up to 5 properties.

Funnels

Conversion across ordered steps, with drop-off and average time between steps.

Retention

Cohort heatmap from a start event and a return event - who comes back, and when.

Segmentation

One aggregate sliced by any dimension, built for KPI tiles and scorecards.

User flow

A Sankey of the paths users actually take between events, where they go next, where they drop.

Top K

Ranked values of any property (top pages, top referrers, top products), with an “others” bucket.

Trends · signups, last 30 days

Funnel · signup → purchase

Retention · weekly cohorts

Cohort W0W1W2W3W4W5W6 Apr 27 100624841363331 May 4 1005845393532 May 11 10064514438 May 18 100614740 May 25 1006652 Jun 1 10063

User flow · after page_view

page_view 100% of sessions, next step →

What product analytics measures Funnel analysis, explained Retention cohorts, explained

Dashboards

The numbers you check every morning

Pin any insight as a tile. The whole board shares one time window, so every tile answers the same question: how is this period doing against the last one?

KPI, line, area, bar, table, and Sankey tiles
One time window and granularity for the whole board
Period-over-period comparison per tile
Thresholds that color a KPI when it crosses a limit
Markdown tiles for notes between charts

Ask your analytics from your editor

Pug serves its read-only analytics API as a Model Context Protocol server. Twelve tools, scoped to one project: an agent runs an insight or looks a person up and gets back the same numbers the dashboard would show you.

Read-only: no tool creates, changes or deletes anything
One project per key, with no way to reach another
The same numbers the dashboard shows you

How the MCP server works

MCP session read-only

You

How are signups trending?

query_insights type: trends · window: 30d

Agent

Signups climbed from 12 a day to 78 across the window, and every dip recovered.

Privacy

Nothing on the device until they say yes

Every default here collects less: nothing is written to a visitor’s device before they choose, and the cloud that receives the events runs in the EU.

Cookieless by default

The SDK starts cookieless: events flow, nothing is written to the device. Identity is derived server-side from a salt that is deleted daily rather than rotated - yesterday’s visitors cannot be reconstructed.

Consent you wire yourself

Point your own banner at three states: granted, cookieless, denied. Activity before consent stays anonymous - nothing is retroactively linked when someone later says yes.

Erasure that hard-deletes

One API call erases a data subject by profile id, or by the external id you assigned. Events, sessions, aliases, and the profile are hard-deleted from ClickHouse and Postgres, and you get back a request id to prove it finished.

Privacy & GDPR How cookieless identity works What it costs your funnels Audit your events for PII

Self-host