Sentry is the category standard and the right default: a free Developer plan (5,000 errors a month), the best stack traces in the category, and a self-hosted version. GlitchTip is the lightweight open-source (MIT) option that accepts Sentry's SDKs. BugSnag and Rollbar are the mature alternatives with stronger release-health framing. Honeybadger bundles error tracking with uptime and cron check-ins, which is the right shape for a solo maintainer. PostHog is the generous free option that attaches a session replay to each error. The thing nobody tells you: error tracking without an alert policy just builds a queue you stop opening. Decide your noise budget before you pick the tool, because that is what you will actually live with.
Error tracking is easy to buy and hard to keep useful. Within a month most installations are generating more notifications than anyone reads.
How we compared
Free tier usefulness at low volume, quality of the stack trace and grouping, self-host option, what alerting looks like by default, and what it costs when a single bad deploy generates fifty thousand events. That last one is where the bills come from. Free allowances and prices were checked on each vendor's own pricing page on October 4, 2026, and are stated as of October 2026 with their billing basis; event-based pricing changes often, so check before you budget.
At a glance
| Tool | Self-host | Free tier (as of October 2026) | Entry paid price | Also does | Watch for |
|---|---|---|---|---|---|
| Sentry | Yes (Fair Source licence) | 5,000 errors/month, 1 user, 50 replays | Team $26/month billed annually, $29 monthly | Tracing, session replay, uptime and cron monitors | Event volume pricing |
| GlitchTip | Yes (MIT) | 1,000 events/month hosted | Small $15/month (100k events) | Uptime checks | Fewer features by design |
| BugSnag | Enterprise on-premises option | 7,500 events/month, 1 user | Select $20/month billed yearly, $23 monthly | Performance monitoring | Priced per event tier |
| Rollbar | No | 5,000 occurrences/month, 1,000 session replays | Essentials from $9/month ($7.50 billed annually) | Session replay | Hard stop at the limit by default |
| Honeybadger | No | 5,000 errors/month, 1 user | Team $26/month ($286/year) | Uptime + cron check-ins | Smaller ecosystem |
| PostHog | Cloud plan compared here | 100,000 exceptions/month | Usage-based, from $0.00037 per exception | Session replay, product analytics | Usage pricing across products |
| Greta | Your deployment | 5 credits a day | Starter $5/month welcome offer (normally $20/month) | Builds the app | Not an error tracker |
Highlight.io, which was on this list, no longer runs as a separate product. LaunchDarkly announced the acquisition in April 2025, Highlight deprecated its services on February 28, 2026 and directed users to LaunchDarkly Observability, and highlight.io now redirects to launchdarkly.com. If you want its successor, LaunchDarkly's free Developer plan lists 5,000 session replays and 5,000 errors a month.
The tools, honestly
Sentry is the default for a reason: the stack traces are the best in the category, grouping is genuinely good, and the SDK exists for whatever you are using. The free Developer plan is one user, 5,000 errors, 50 replays and 30 days of history; Team is $26 a month billed annually ($29 monthly) with unlimited users and 50,000 errors. It is also the one most likely to surprise you on cost, because a single crash loop can emit tens of thousands of events in an hour. Turn on spike protection, which applies a dynamic rate limit per project, and set a per-key rate limit on day one, not after the invoice. Self-hosting is free, but note the licence: Sentry describes itself as Fair Source under the Functional Source License, which is not OSI open source, and the self-hosted version has no spike protection and no dedicated support.
GlitchTip says it is "compatible with Sentry client SDKs", so you can point an existing Sentry SDK at it, and it is MIT-licensed if you want to self-host. The hosted free plan is 1,000 events a month (uptime checks count as events), and Small is $15 a month for 100,000. That is a real option for a small app where the hosted Sentry bill is disproportionate, and the trade is a smaller feature surface.
BugSnag and Rollbar both frame around release health (is this version worse than the last one), which is a more useful question than raw error count. BugSnag is owned by SmartBear; its free plan is one user and 7,500 events a month, and Select starts at $20 a month billed yearly ($23 monthly). Rollbar's free plan gives 5,000 occurrences and 1,000 session replays a month with unlimited users, and paid plans are Essentials (from $9 a month) and Advanced (from $13 a month). Honeybadger is the best-shaped for a solo maintainer: errors, uptime and cron check-ins in one bill, and its free Developer plan includes 5,000 errors a month plus one monitor (uptime check or check-in), so you are not assembling three tools to answer "is it up and did the nightly job run". PostHog puts the session replay next to the error, which turns "cannot reproduce" into a recording.
Got an idea? Build it now!
Just start with a simple prompt. No coding required: Greta.sh turns your idea into a working app in minutes.
Best free error tracking tools
If the budget is zero, the free tiers are not equal. PostHog is the largest by volume (100,000 exceptions a month on its pricing page, with 5,000 session recordings). BugSnag gives 7,500 events, while Sentry, Rollbar and Honeybadger each give 5,000. Sentry and Honeybadger limit the free plan to one user; Rollbar's free plan has unlimited users. GlitchTip's hosted free plan is the smallest at 1,000 events, but its MIT-licensed self-hosted version costs nothing beyond the server you run it on. Pick on the volume your app actually produces in a bad week, not a good one.
Production error tracking: what to set up
Installing the SDK is ten minutes. Making it useful in production is three settings: tag every event with a release (so you can tell whether a deploy made things worse), upload source maps or debug symbols (or your stack traces point at minified code), and set a rate limit before the first crash loop rather than after it. The alert policy below is the fourth.
Session-based error monitoring
Some tools attach a recording of what the user did to the error itself. Sentry links errors, traces and console logs to the replay timeline (50 replays a month on the free plan). PostHog says "every issue comes with the affected user's session replay" (5,000 recordings free). Rollbar's free plan includes 1,000 session replays. LogRocket is replay-first, with errors unlimited on every plan and 1,000 sessions on its free plan. GlitchTip, BugSnag and Honeybadger do not list session replay on the pricing pages we checked. Replay matters most for front-end bugs that only happen in a specific click sequence; for a backend job that fails at 3am, the stack trace is the evidence.
Got an idea? Build it now!
Just start with a simple prompt. No coding required: Greta.sh turns your idea into a working app in minutes.
The part that decides whether you keep using it
Alert policy. The default configuration on most of these notifies on every new issue, which is correct for week one and unbearable by week four. A workable policy for a small app: alert immediately on a new issue affecting more than one user, alert on any regression of a previously resolved issue, and digest everything else daily. Everything that does not meet the bar goes to a dashboard you check deliberately, not to a channel you mute.
The second thing: attach a release identifier to every event. Without it you cannot answer whether a deploy made things worse, which is the only error-tracking question that reliably changes what you do next.
Where this sits if you built the app with AI
An app generated from a description still throws exceptions at runtime, and the tracking story is identical: install an SDK, tag releases, set alert rules. Worth being plain: Greta builds and deploys the app, it does not replace an error tracker, and you should install one of the six above. What it does change is the fix loop, because a good stack trace plus the bug report template is exactly the input needed to describe a fix precisely. The free tier is 5 credits a day; Starter is a $5/month welcome offer (normally $20/month).
In the weeks right after launch, reading every new error each day is one line of a first-month operating plan, next to a restore you've actually tested, an uptime check and shipping fixes one at a time so each new error can be traced to a release. Error tracking tells you something broke inside the app; an uptime monitor tells you nobody can reach it, and you want both. If a webhook handler is the thing throwing, the webhook tools guide covers replaying the events it dropped, and when an error turns into an outage, the incident handoff template keeps the fix and the follow-up in one place.
FAQ
What is the best error tracking tool? Sentry for most people. GlitchTip if you want to self-host cheaply and your SDK already speaks Sentry. Honeybadger if you want uptime and cron monitoring in the same bill. PostHog if reproduction is your bottleneck and you want replay attached to each error.
What is the best free error tracking tool? For volume, PostHog (100,000 exceptions a month free). For the best stack traces on a free plan, Sentry's Developer plan (5,000 errors a month, one user). For a team on a free plan, Rollbar (5,000 occurrences a month, unlimited users). Sentry, BugSnag, Rollbar, Honeybadger, GlitchTip and PostHog all have a free tier; Raygun does not (it offers a 14-day trial).
Is Sentry's free tier enough? For a small app with a healthy error rate and one developer, often yes. It stops being enough exactly when you need it most, during an incident generating a flood of events, so turn on spike protection and rate limits regardless of tier.
Is Sentry open source? Not in the OSI sense. Sentry describes itself as Fair Source under the Functional Source License, which converts to Apache 2.0 after two years. If you need an MIT-licensed tracker that accepts Sentry SDKs, GlitchTip is the usual choice.
What happened to Highlight.io? LaunchDarkly acquired Highlight in April 2025, and Highlight deprecated its services on February 28, 2026. Former users were directed to LaunchDarkly Observability.
Do I need error tracking and uptime monitoring? They answer different questions. Error tracking says something broke inside the app; uptime monitoring says nobody can reach it. An app can be fully up and broken for every user, and down with zero errors reported.
How do I stop alert fatigue? Raise the bar for immediate alerts to things affecting more than one user, alert on regressions, digest the rest. An alert channel everyone has muted is worse than no alerting, because it looks like coverage.
