PostHog VS Sentry

This is the same model that PostHog, Supabase, and dozens of other developer tools use. Open core, with a managed offering on top.

Offchain: Website traffic, in-app behaviour, marketing channels, growth campaigns (Google Analytics or PostHog).

Topic PostHog (Web Vitals) Apogee Watcher Primary job Product analytics OS; Web Vitals are real-user metrics from the browser Synthetic PageSpeed monitoring + CrUX in results Instrumentation Requires posthog-js on the site No script on monitored sites Metrics FCP, LCP, INP, CLS from real sessions ($web\_vitals) when capture runs Lighthouse lab + CrUX (where available) via PSI Cookieless analytics With always (and cookieless paths without session IDs), $web\_vitals does not populate—see Cookieless mode above No snippet; tests do not use PostHog session state Best for “How do my users experience my app?” “How are these URLs doing on a schedule—and across many clients?” Multi-site agency Analytics projects and teams—not Watcher’s org/site/page model Multi-tenant orgs, roles, discovery, budgets Budgets & email alerts Insight-based—build trends from $web\_vitals, attach [alerts](https://posthog.com/docs/alerts) with thresholds, frequency, destinations; maintain as analytics evolves Monitoring-native—performance budgets and email alerts tied to scheduled tests; no separate insight to curate first Extras Flags, replay, experiments, cohorts, warehouse pipelines PDF-style reporting direction, Leads prospecting workflows Cost shape Event-based (vitals count toward event quotas) Plan-based subscription; PSI quota bundled—verify [pricing](https://apogeewatcher.com/pricing)

PostHog provides an open-source analytics approach, giving teams flexibility to customize dashboards, session recordings, and heatmaps.

Originally built for Web2, PostHog can be extended with Web3 event tracking through custom data pipelines. Key features of Posthog are funnels, session recordings, feature flags, open-source & self-hosted

A big thanks to PostHog who pioneered this approach with HogQL, a SQL-like interface on top of ClickHouse. TRQL started as a TypeScript conversion of their Python implementation but evolved significantly during development to handle our specific use cases.

Plausible brought open source to web analytics. Cal.com did it for scheduling. Formbricks did it for surveys. PostHog did it for product analytics. Quackback does it for feedback collection.

Is Sentry (or equivalent error tracking) wired up from day one?

Sentry captures every error in your application with full stack traces, breadcrumbs (the sequence of events leading to the error), and user context. It works across every major platform: JavaScript, Python, Go, Ruby, Java, .NET, iOS, Android, and more.

Sentry has long been considered the gold standard of error tracking, and their free tier in 2026 is remarkably generous. With 5,000 errors per month and 1 GB of storage, it's perfect for small to medium projects and even some larger applications with good error hygiene.

For these cases, use AI as one tool alongside traditional debugging: breakpoints, profilers, log analysis tools like Sentry or Datadog, and good old rubber duck debugging. When bugs escalate into production outages, having an AI incident management workflow in place helps your team respond systematically instead of scrambling.

Sentry is indeed, and is open source and self-hostable: https://github.com/getsentry/sentry

It’s a pretty large app (I think >1M lines of Python). I ran a self hosted version for several years and found it performant and pretty easy to keep running and updated.

Sentry – error tracking with real context

Error monitoring is configured (Sentry, Rollbar, or similar).

Tembo integrates with observability tools like Sentry, Datadog, and New Relic to detect performance regressions, query slowdowns, and recurring exceptions. When it finds issues, it analyzes recent commits, traces the root cause, and opens pull requests with fixes or optimizations — often before anyone notices the problem.

This way, the primitive allows you to choose the layer in your application for throwing and catching. One of our key ideas was to define integration with Sentry logging "at the top":