Shipping

Error Monitoring

Find out when your app throws an error in production, with the stack trace and context to fix it, before users email you.

Ask your AI about this, with this page as the source:ChatGPT ↗Claude ↗Perplexity ↗

The real choice

Hosted service or self-hosted tracker. Hosted services give you setup in minutes and pricing by event volume; self-hosting keeps error data on your servers for the cost of running and upgrading it.

Pick by situation

Tap the ones that are you — the tools that fit light up below.

You want the widest SDK coverage and tracing next to errorsSentry
Your product is a mobile app and crash-free rates per release matterSentry
Your error data must stay on your own serversGlitchTip Sentry

The contenders

Grouped by the side of the choice they answer, not ranked. Open a row for when to use it, the trade-off and what makers say.

ToolBest for

Hosted service2

Sentryservice · self-hostable · free tierBroad SDK coverage across languages and frameworks, with performance tracing next to errors.

162+263 open source$29Team33.8M/wk+4% vs npm

Use it whenYou want one tool that covers every language and framework in your stack, frontend to backend.

Trade-offEvent-based billing can jump after a noisy bug; the self-hosted version needs a sizeable server.

llms.txt MCP CLI

Used by 24CallDesk, AgentQL, Beatoven.ai, bestcourses, Brev.io and 157 more · in 263 open-source projects.

Sentry in full →Plans and calculator →vs Rollbar →vs GlitchTip →

Sentry gives us one place to investigate errors across Marqly’s web app and API. The error context is especially useful when tracing a failed save, sync, or AI Organizer run.

Marqly, the makerSep 2026 ↗

Rollbarservice · free tierError grouping tied to releases, with session replays linked to errors.

6+16 open source$29Essentials 50k498.7k/wk−59% vs npm

Use it whenYou want each error tied to the deploy that introduced it.

Trade-offSmaller ecosystem of integrations and guides than Sentry.

llms.txt

Used by Hovercode, CircleCI, Cloudinary, Frill, Shortcut and 1 more · in 16 open-source projects.

Rollbar in full →Plans and calculator →vs Sentry →

Self-hosted tracker1

GlitchTipservice · open source · free tierSelf-hosting error tracking on a small server while using Sentry's SDKs.No evidence yet$15Small

Use it whenYou want error data on your own infrastructure or a lighter stack than self-hosted Sentry.

Trade-offFewer features than Sentry, and you run the upgrades and backups.

CLI

No maker's product we track shows it for this yet.

GlitchTip in full →Plans and calculator →vs Sentry →

Cost: the cheapest plan that fits errors 50k events, from list prices. Try your own numbers →

Cost as you grow

Each contender's cheapest usable plan as usage rises.

GlitchTip Rollbar Sentryx: errors per month (event) · cheapest usable plan at each point, list prices · try your own numbers

Who switches to what

Public pull requests on GitHub since Oct 2024 whose title says "X to Y" — real code changes, by developers in general rather than makers only. Pick a flow to see its pull requests.

Sentry →

GlitchTip4 PRs · Compare →

ADFA-4751: Rebrand remaining Sentry references to GlitchTipappdevforall/CodeOnTheGo · 2026-07-17
feat(telemetry): guarded /api/debug/sentry route to verify GlitchTip captureKhourySpecialProjects/video-review-system · 2026-07-17
ADFA-4397: Migrate crash and log reporting from Sentry to GlitchTipappdevforall/CodeOnTheGo · 2026-07-07
feat(observability): wire errors-only Sentry SDK to GlitchTipsolid-stats/server-2 · 2026-06-14

All matching pull requests on GitHub ↗

Before you choose

How to approach it

Install the SDK on both frontend and backend before launch, upload source maps, and tag each release so you can see which deploy introduced an error. Set alerts for new issues only, not every occurrence, or you will start ignoring them. Filter out noise (bots, browser extensions) early to stay inside your event quota.

Common mistakes

Shipping minified frontend code without uploading source maps, so every stack trace points at one unreadable line.
Alerting on every occurrence instead of new issues, until the channel is muted and real regressions go unseen.
Sending request bodies and headers with passwords, tokens or personal data into a third-party error tracker.