GlitchTip vs Raygun: A Hands-On Error Tracking Comparison for 2026

Stanley Ulili
Updated on November 19, 2026

Count the subscriptions before you count the features. GlitchTip sells one plan that covers errors, logs, route timing and uptime checks from a single pool of events, and you can skip the bill entirely by hosting it yourself. Raygun is three subscriptions in one account, one each for crashes, real users and server timing, and taking all three at the entry tier costs about thirteen times GlitchTip's smallest plan.

What the extra money buys turned out to be specific. GlitchTip is open-source error tracking that takes data from Sentry's SDKs, so you instrument your app with Sentry's SDKs and get source lines, logs and route averages on top. In my test it caught every failure I planted and showed the failing server code without any repository connected.

Raygun is a monitoring suite whose browser product measured page speed and Web Vitals with no extra code, and whose deployment page named the bug I'd shipped again. Its server performance agent, on the other hand, saw none of my code and broke its own error reports.

Both tools watched the same storefront and API through three releases. The sections below compare setup, capture, debugging, source maps, grouping, releases, the browser, server performance, logs, alerts, AI and cost.

GlitchTip vs Raygun: Results at a glance

My calls rest on a single Express API and Vite storefront, wired to each vendor by its own guide. Another language, or Raygun's paid upload token, might move a handful.

Topic Who did better Reason
Setup GlitchTip Errors arrived on the first request; Raygun needs an application per tier and a paid token for automation
Error capture GlitchTip It caught the unhandled rejection Raygun missed
Debugging an error GlitchTip Server source lines with no repository; Raygun shows frames only
Source maps GlitchTip Debug IDs from sentry-cli; Raygun needed hand-labelled uploads on the trial
Error grouping Neither Both split one payment bug by the amount in its message
Releases and regressions Raygun Each deployment page lists its regressions; GlitchTip's release page shows files
Browser monitoring Raygun Vitals, route timings and a page per shopper; GlitchTip records no page loads
Server performance GlitchTip, barely A route average rose; Raygun's APM saw no app code and emptied error reports
Logs GlitchTip A searchable log list; Raygun stores no logs
Uptime and alerts Not compared No alert was delivered; GlitchTip's uptime checks weren't tested
AI-assisted debugging Not compared I left both AI features untried
Price GlitchTip A free tier, 100,000 events for $15, or self-host; Raygun starts at $40 per product

How I tested GlitchTip and Raygun

monitoring-lab, the shop I break for every review, runs an Express API on SQLite beside a mock price service that fails when told to and a sales report job that runs each minute. Its Vite storefront routes by hash. Swapping vendors touches two small adapter files and nothing else.

Planted failures:

  • In the storefront: the wishlist button reads a property of null, order tracking rejects without a handler, and some API calls come back 404, 500 or 502.
  • API side: an unknown coupon crashes checkout, requesting a nonexistent book crashes that route, sold-out stock is reported as a handled error, and the price service sometimes answers 503.
  • Async: a recommendations promise rejects with nothing listening, and the report job fails on some runs.

Traffic comes from two headless Chromium shoppers, alice and bob, and from carol, who calls the API with no browser.

Release plan

Version Checks
1.0.0 First events, source maps, what an error shows
1.1.0 The coupon bug set as fixed for the next version, then redeployed as is
1.2.0 The checkout's book lookup slows to seconds, and one card charge in five fails

1.2.0 is the hardest test. Its slowdown never throws, so only a tool that times requests can see it.

SDKs and coverage

Platform GlitchTip Raygun
API @sentry/node 11.8.0 raygun 2.2.10, plus raygun-apm 2.3.0 and the APM agent
Storefront @sentry/browser 11.8.0 raygun4js 3.1.4
Plan Hosted GlitchTip, Free plan 14-day trial of Crash Reporting, RUM and APM
  • GlitchTip's free cap: I sent 1,097 events against a 1,000-event plan, so its final performance figures may be incomplete.
  • Raygun automation: source map and deployment uploads by API need a paid-only token, so on the trial I did both by hand.
  • Raygun APM in 1.2.0: the APM package had to be switched off for that run, for reasons covered under performance, so there's no APM data for the slow release.
  • Not tested: GlitchTip's uptime checks (the app wasn't reachable from outside), alert delivery and both tools' AI features.

Terms in each tool

Idea GlitchTip Raygun
Grouped errors Issue Error group
One occurrence Event Instance
A part of your app Project Application
Release release from the SDK Version from setVersion, plus deployments
Fixed for a version Resolve in next release Resolve in next version

Setup and onboarding

GlitchTip's trap hides your performance data. Raygun's traps slow down everything else.

GlitchTip: Plan first, then a copy of Sentry's setup

You pick a GlitchTip plan before you can make a project, and its project form wouldn't save until I'd added a team with the + beside that field. The connect page that follows is Sentry's Node.js code with your DSN filled in:

GlitchTip connect page with the Sentry SDK snippet

The snippet sets autoSessionTracking: false, with a comment noting that GlitchTip has no sessions. Errors appeared from the first request. Performance didn't: version 11 of Sentry's SDK sends spans as a stream, a format GlitchTip doesn't read, since it expects transactions:

GlitchTip Performance page reporting no transactions

The page gives no hint why. Adding traceLifecycle: 'static' to the SDK options fixed it, and the connect page doesn't mention it.

Raygun: One application per tier, one product at a time

Raygun starts by creating an application:

Raygun onboarding step to create the first application

Each application covers one part of the app, with its own API key and error list, so I made monitoring-lab-api and monitoring-lab-web. The wizard sets up one product per pass and won't let you finish until a first event lands. Three more snags:

  • raygun4js from npm stayed silent in a lazily loaded module. It reads its settings on the page's load event, so a module loaded later never sent anything until I queued calls the way the CDN snippet does.
  • API uploads need a Personal Access Token, which only paid plans get. On the trial, source maps and deployments were manual.
  • APM runs as a separate agent beside your app, in my case a Docker container.

Worth turning on: in GlitchTip with Sentry SDK 11, set traceLifecycle: 'static' before your first deploy, or Performance stays blank with no explanation. In Raygun, put the raygun4js setup in the page's first script if you install from npm.

Setup comparison

Step GlitchTip Raygun
Browser and API together ✔: one project ✘: one application each
Install Sentry's SDK and your DSN Raygun's SDKs, one product per pass
First errors ✔: at once ✔: at once on the server
Hidden catch Blank Performance on SDK 11 npm browser build silent when loaded late
Upload automation on the free tier or trial ✔: sentry-cli ✘: paid token needed

Verdict: GlitchTip. Its catch only touched performance, and errors worked immediately on both tiers. Raygun's split into applications and products turned a small app into several setups.

Error capture

Both tools received the same failures on both tiers.

GlitchTip: All of it

After one traffic run, GlitchTip held 24 issues:

GlitchTip issues list after the first run

Every planted failure is there, with 6 events on the unawaited recommendations promise. The list is longer than my bug count for reasons covered under grouping, not capture. Nothing marks a row as handled or unhandled.

Raygun: Five of six on the API

Raygun's API application showed five groups:

Raygun error groups for the API application

The two TypeErrors, the sold-out error, the 503 and the report job each have a group. The recommendations rejection never arrived, with reportUncaughtExceptions switched on. Browser errors went to the web application, covered under the browser.

Error capture comparison

Test GlitchTip Raygun
Wishlist crash ✔: received ✔: received
Coupon and book crashes ✔: received ✔: received
Errors the app reported ✔: received ✔: received
Rejected promise, no handler ✔: separate issue ✘: never arrived
Report job ✔: caught ✔: caught
Both tiers in one list ✔: one project ✘: two applications

Verdict: GlitchTip. It caught everything, the rejection included. If you rely on Raygun for Node.js, test your unhandled-rejection path first.

Debugging an error

The coupon crash, as each tool shows it.

GlitchTip: The server's code on the page

GlitchTip displayed the failing lines straight from the SDK's payload:

GlitchTip server error with source lines around line 150

Lines 145 to 155 appear, including my comment marking the planted bug, because Sentry's SDK sends context lines itself. Its breadcrumb trail was less tidy. Thirteen came with the coupon error, yet only two were from that checkout, with startup lines and report-job chatter filling the rest.

Raygun: Tidy breadcrumbs, no code

Raygun's error page opens on totals, affected users, first and last seen, and a chart of occurrences:

Raygun API error page with instances and affected users

Tabs hold the summary, details, HTTP, environment, custom data, breadcrumbs and raw payload, and the affected-user count links to the Customers page. Server breadcrumbs stayed inside the one request:

Raygun breadcrumbs for one checkout request

The request, the start of checkout and the book lookup appear, and nothing from other shoppers. The trace gave a file and line number plus one folded framework frame, and no code.

Debugging comparison

Capability GlitchTip Raygun
Server source lines ✔: from SDK context ✘: frames only
Breadcrumbs Mixed with other requests ✔: scoped to the request
Affected users ✔: on the event ✔: linked to Customers
Raw payload ✔: raw JSON ✔: Raw data tab
Browser error linked to server error ✘: separate issues ✘: separate applications

Verdict: GlitchTip. Reading the failing line on the error page saved the most time. Raygun's request-scoped breadcrumbs are the cleaner trail.

Source maps

Maps had to be uploaded in both, because my host isn't reachable from outside.

GlitchTip: Debug IDs, one command

sentry-cli injects a debug ID into each bundle and uploads its map, and GlitchTip matches them without any URL setup. The wishlist crash resolved to ../../src/pages.js line 122 on the first try.

Raygun: Labelled by hand on the trial

Uploads by API need the paid token, so I used the JS Source Map Center:

Raygun JS Source Map Center with the paid-token note for API uploads

Every file needs its absolute URL typed in as a label, and the names change on every build. Once that was done, "Re-process this error for Source Maps" mapped the crash at once:

Raygun browser error mapped to pages.js line 122

The frame shows ../../src/pages.js:122 with the lines around it.

Capability GlitchTip Raygun
Matching ✔: debug IDs URL labels
Upload on the free tier or trial ✔: CLI Manual, per build
Re-map past errors Not tested ✔: re-process button

Verdict: GlitchTip. Both mapped correctly. GlitchTip's upload fits into a build script, and Raygun's needed hand work after every build until you pay.

Error grouping

1.2.0 added a payment error whose message carries the charge amount, which tests any grouping rule built on message text. Both tools build on it.

GlitchTip: One issue per amount

GlitchTip groups by message, so each amount opened a new issue. One bug became seven:

GlitchTip payment gateway error split into one issue per amount

Every amount, 35.99 USD at the bottom to 84 USD at the top, holds a single event in its own issue. Failing report runs split by run number in the same way. Old bugs held steady through the 1.2.0 code change and simply grew.

Raygun: Split by amount and by URL

Raygun's Node.js grouping also reads the message, so the payment error became two groups sharing a class and a stack. In the storefront, one failed-load error split by URL:

Raygun browser error groups split by URL

/books/0, /404, /99 and /1234 each got a group, while order tracking held one group across changing IDs. On top of that, 1.2.0's code change gave two older server bugs a second group each.

Grouping comparison

Test GlitchTip Raygun
Two TypeErrors at 1.0.0 ✔: kept apart ✔: kept apart
New payment error ✘: 7 issues ✘: 2 groups
Same error, different URLs ✘: split by query string ✘: split by path
Old bugs after the code change ✔: grew in place ✘: second groups opened
Fixing it Merge by hand Not tested

Verdict: Neither. Both let values in a message decide the group. GlitchTip split harder on amounts, and Raygun split old bugs when the code moved.

GlitchTip opened seven issues for one payment bug, and Raygun opened two plus duplicates of older bugs. Better Stack, fed by the same Sentry SDK GlitchTip uses, kept that bug in one group:

Better Stack single payment error group with source

Ten exceptions with ten amounts sit together, with the source around the throw. Older bugs kept their single rows through the release.

Keep one bug in one group

Releases and regressions

For 1.1.0, each tool was told the coupon crash would be fixed in the next version. Then the unchanged code shipped.

Raygun: A Regressions tab per deployment

Raygun's resolve dialog has an option GlitchTip lacks:

Raygun Resolve in next version dialog

Its discard checkbox comes ticked and throws away further 1.0.0 occurrences until the next version, which keeps them off your quota. On 1.1.0 the coupon group went Active with "Error reoccurred" in its activity, and its list row simply read Active. The deployment page made the difference:

Raygun deployment 1.1.0 with one regression

Its Regressions tab holds one group, the coupon TypeError, resolved in 1.0.0 and active again. On the trial, each deployment was made by hand, and the first save hit a Raygun 500 page until I left every field but Version blank.

GlitchTip: Reopened, with a release page of files

GlitchTip offers "Resolve in next release", and the coupon bug came back on 1.1.0:

GlitchTip coupon error back to Unresolved on 1.1.0

It's back to "Unresolved", with the newest event tagged 1.1.0. No regression label appears or reopen history, and the release page lists uploaded files, not errors. Finding what 1.2.0 introduced meant sorting the issue list by first seen.

Release comparison

Capability GlitchTip Raygun
Resolve for a future version ✔: next release ✔: next or a chosen version
Returning bug reopened ✔: yes ✔: "Error reoccurred"
Regression label in the list ✘: none ✘: none
Errors per release ✘: release page lists files ✔: deployment page with Regressions
Deployments From the SDK release Manual on the trial, API with a paid token
Drop repeats from the old version ✘: no ✔: discard option

Verdict: Raygun. Its deployment page answers "did my fix hold?" in one tab. GlitchTip reopens the bug and leaves the release questions to you.

Both tools reopened the coupon bug, and neither marked it in the main list. Better Stack changed the row itself when 1.1.0 arrived:

Better Stack list with the coupon bug marked Reoccurred

A red "Reoccurred" sits on the coupon row alone, taken from the SDK's release value with no deployment to create. The failed fix stands out in the list you read every day.

Notice a fix that didn't hold

Browser monitoring

Raygun sells browser monitoring as a product. GlitchTip records browser errors and little else.

Raygun: Page speed from the script alone

Real User Monitoring started measuring as soon as raygun4js loaded:

Raygun RUM performance charts

Load time sits beside the Core Web Vitals (LCP, CLS, INP) and first paint. Timings per route came from its virtual pages table (/books/* 1.58 s, /checkout 1.76 s), and API calls had their own table. Each shopper gets a Customers entry tying sessions to errors:

Raygun Customers page with sessions and errors per user

alice had 10 sessions and 5 errors, and each session opens as a timeline of pages and time spent. There's no video replay. Browser breadcrumbs were automatic, covering clicks, hash changes and requests.

GlitchTip: Errors with browser context

A storefront error in GlitchTip carries who hit it, plus browser, OS, release and environment. No page load ever came through as a transaction, despite browser tracing and the static setting, and replay isn't offered.

Capability GlitchTip Raygun
Browser errors ✔: in the shared project ✔: in the web application
Web Vitals ✘: none recorded ✔: LCP, CLS, INP
Speed by route ✘: none recorded ✔: virtual pages
Per-user view ✘: not offered ✔: Customers
Session replay ✘: not offered ✘: timeline only
Browser error linked to server error ✘: no ✘: no

Verdict: Raygun. Beyond errors, nothing in this test was more useful than its RUM, which needed no code past the error script.

GlitchTip never timed a page, and Raygun's page timings stop at the browser, with the API in another application. Better Stack drew one visit across both tiers:

Better Stack timeline joining storefront and API spans

Each click and fetch sits above the API spans it caused, from one SDK setup. A slow step shows up wherever it happened.

Follow a visit from the click to the API

Server performance

1.2.0 slowed checkout from milliseconds to seconds without an error. Neither tool caught it per release.

GlitchTip: A rising average

With the static lifecycle set, GlitchTip ranked routes by average duration:

GlitchTip transaction groups after 1.2.0

Checkout's average reached 752.27 ms, blended across all three releases, with no release filter to separate 1.2.0. Route pages show aggregates and a span breakdown, but no single requests to open.

Raygun: An agent that saw nothing, and broke errors

Raygun APM needed raygun-apm and its agent. Locally, nothing reached the agent until RAYGUN_AGENT_HOST pointed at 127.0.0.1, which the Node.js setup page never mentions. The overview then filled in:

Raygun APM overview with Apdex and request speeds

It showed Apdex 0.94, an 83 ms median and 4,754 ms at p99, with queries per minute stuck at zero because the profiler doesn't support SQLite. A 5.3-second trace broke down into garbage collection and idle time, with none of my functions in it.

Worse, installing the package changed error reports:

Raygun errors arriving as Unknown error after installing APM

Once raygun-apm was in the project, native errors reached Raygun as blank "Unknown error" reports, profiler on or off. Crash Reporting pulls in a bridge from the APM package, the bridge swaps out global.Error, and native errors then fail the instanceof Error check. Removing the package, or the environment variable that disables the bridge, restores them, and that's why 1.2.0 ran without APM. Raygun's RUM release comparison called 1.2.0 34.8% faster, since it measures page loads, not the slow checkout request.

Performance comparison

Capability GlitchTip Raygun
Setup traceLifecycle: 'static' A separate agent and package
App code in traces Span breakdown per route ✘: none seen
SQLite queries Not checked ✘: not supported
Release filter ✘: none ✔: RUM compare, page loads only
Effect on errors None ✘: emptied native errors until disabled
1.2.0 slowdown Partly: blended average rose ✘: missed

Verdict: GlitchTip, barely. A blended average that rises is a weak signal, but it's a signal. Raygun's APM cost me error data and saw none of my code.

GlitchTip offered an average across releases, and Raygun's agent missed the slow checkout entirely. One failing 1.2.0 checkout in Better Stack opened into its spans:

Better Stack spans for the slow 1.2.0 checkout

The book query ate 1.53 s of that 2.19 s request, and the refused charge 660 ms. Its logs needed one filter:

Better Stack Live tail filtered to one failing checkout

The app's slow-query warning sits right before the failure, from the same SDK, with no agent to run.

Trace a slow release to its query

Logs

GlitchTip stores logs through the Sentry SDK. Raygun stores none.

GlitchTip kept 233 log lines from the first run:

GlitchTip Logs with level and service filters

Each row has a time, level and message, with text search and filters for level and service. A line opens nothing further, and error pages don't link to their logs. Raygun has no log product; you'd pair it with a separate tool.

Capability GlitchTip Raygun
Keeps app logs ✔: sent by Sentry's SDK ✘: not offered
Find a line ✔: search plus level and service filters ✘: not offered
Logs linked to errors ✘: no ✘: not offered
Billing Ten lines count as one event Not applicable

Verdict: GlitchTip. It keeps your logs searchable, if unlinked. Raygun leaves logs to another vendor.

Uptime and alerts

I connected no notification channel, so nothing was delivered.

GlitchTip includes uptime monitors that draw from the same event pool as errors and logs. They probe from GlitchTip's side and couldn't reach my app, so I didn't test them.

Raygun gives each new application its own notification defaults:

Raygun Create application form with notification defaults

Out of the box you get mail for new errors, a digest each day and a RUM report each week; APM alerts begin disabled. Triage options are Resolve, Ignore, Permanently ignore and the version-based resolves. I found no uptime checks in Raygun.

Capability GlitchTip Raygun
Uptime checks Included, not tested ✘: not found
New-error emails Not tested ✔: on by default
Digests Not tested ✔: daily, plus weekly RUM
Alert delivered ✘: not tested ✘: not tested

Verdict: Not compared. Raygun's notification defaults are richer, and only GlitchTip watches uptime. Neither fired.

AI-assisted debugging

I didn't run either tool's AI.

  • GlitchTip: a "copy for AI" link beside "raw JSON" copies the event for your own assistant.
  • Raygun: AI Error Resolution, for likely causes and fixes, is listed under Crash Reporting at every tier, and a remote MCP server lets coding assistants read your data.
Capability GlitchTip Raygun
In-app analysis ✘: not offered AI Error Resolution, not tested
Hand-off to your assistant "copy for AI" Remote MCP server, not tested
Cost Free Included in Crash Reporting

Verdict: Not compared. Raygun has more to offer on paper, and I tested neither.

GlitchTip leaves AI to your clipboard, and Raygun's features went untried here. Without a repository, I asked Better Stack's AI SRE what 1.2.0 broke:

Better Stack AI SRE answer on release 1.2.0

It isolated the book lookup as the one statement that slowed and tied 10 failed charges out of 48 to the new release. The run took around four minutes and cost $0.49.

Have AI SRE explain a deploy

Pricing

GlitchTip draws errors, transactions, logs and uptime checks from one event pool. Raygun bills Crash Reporting, RUM and APM as separate subscriptions.

GlitchTip's prices were read on November 4, 2026, Raygun's on October 30. Confirm them before committing.

Starting out

Item GlitchTip Raygun
Free plan ✔: 1,000 events a month ✘: none listed
Free retention 30 days No free plan
Trial Not needed 14 days, no card
Past the limit Some events dropped, all blocked at 2x Per-unit charges or a bigger tier
Self-hosting ✔: open source ✘: hosted only

GlitchTip's free tier fills quickly once performance data is on, since every transaction counts:

GlitchTip usage at 110% of the free limit

My test ended at 110% of the limit with a warning banner, and events kept arriving.

Entry-level paid plans

Entry tier GlitchTip Small Crash Reporting Basic RUM Basic APM Basic
Paid yearly Not listed $40 per month $80 per month $80 per month
Paid monthly $15 $60 per month $120 per month $120 per month
Included 100,000 events 100,000 errors 100,000 sessions 100,000 traces
Data kept 90 days 180 days Not listed Not listed
Users Not limited on the page Unlimited Unlimited Unlimited

Above Small, GlitchTip sells half a million events at $50 and 3 million at $250 monthly, with overage from $0.17 per thousand under a cap you set. Raygun bills overage at $0.001 an error and $0.002 a session or trace, and its Crash Reporting plan bundles a starter allowance of RUM sessions and APM traces.

Comparing the actual cost

For errors alone, GlitchTip's $15 buys the same 100,000 volume as Raygun's $40 to $60 Crash Reporting, and adds logs, route timing and uptime from the same pool. Raygun keeps errors twice as long, at 180 days.

The gap widens with the browser. Raygun's three Basic products come to $200 a month on annual billing, and RUM is the part I'd actually pay for. GlitchTip has no RUM, so matching that means a second tool. Our GlitchTip vs Rollbar comparison sets GlitchTip against a tracker whose entry tier is cheaper still.

Verdict: GlitchTip. It's free to start, cheapest at volume and free to self-host. Raygun costs more for errors, and its browser monitoring is the part worth the money.

Full feature comparison

Feature GlitchTip Raygun
SDKs Sentry's Raygun's own
Browser and server together ✔: one project ✘: one application each
Unhandled rejection ✔: caught ✘: missed in my setup
Server source lines ✔: from the SDK ✘: frames only
Request-scoped breadcrumbs ✘: mixed ✔: yes
Source map upload on the free tier or trial ✔: CLI ✘: manual
Grouping on message values ✘: splits ✘: splits
Regressions per release ✘: not offered ✔: Regressions tab
Web Vitals and route speed ✘: not recorded ✔: RUM
Per-user sessions ✘: not offered ✔: Customers
Session replay ✘: not offered ✘: timeline only
Server performance Route averages, no release filter APM agent, limited in my test
Logs ✔: searchable list ✘: not offered
AI "copy for AI" AI Error Resolution and MCP, not tested
Uptime checks Included, not tested ✘: not found
Free plan ✔: 1,000 events ✘: none
Self-host ✔: open source ✘: hosted only

Which should you choose?

GlitchTip gives you the error and the code cheaply. Raygun gives you the user's experience, at a price.

Choose GlitchTip if you want Sentry-compatible errors without the bill

  • You want the failing server line on the error page. It came from the SDK with no repository.
  • You want logs, route timing and uptime in one plan. All draw from one pool, from $15 a month, or free on your own servers.
  • You might switch later. Its Sentry compatibility makes moving a matter of changing the DSN.

The costs: a group for every value in a message, mixed breadcrumbs, a release page that lists files, no browser performance and a blank Performance page on SDK 11 until you change one setting. Our Sentry vs GlitchTip comparison puts it beside the platform whose SDKs it uses.

Choose Raygun if browser experience is what you measure

  • You care how fast pages feel. RUM gave Web Vitals, speed per route and per-user sessions with no extra code.
  • You review each deployment. The Regressions tab names the fix that didn't hold.
  • You want long retention. Crash Reporting keeps errors for 180 days.

What you accept: an application per tier, a missed unhandled rejection, manual uploads until you pay for a token, server stacks without code, and an APM package that emptied my error reports. Our BugSnag vs Raygun comparison weighs Raygun against a tracker focused on release health.

FAQ

Does GlitchTip work with Sentry SDK 11?

It does for errors and logs. For performance, add traceLifecycle: 'static', since SDK 11 streams spans and GlitchTip only understands transactions.

Is it safe to add Raygun APM to a Node.js app using Raygun Crash Reporting?

Test it first. With raygun-apm 2.3.0 beside raygun 2.2.10, every native error arrived as an empty "Unknown error" until I set RAYGUN_APM_CRASH_REPORTING_INTEGRATION=false.

Does either tool have session replay?

Neither records video. Raygun lists each session's pages and time spent; GlitchTip keeps no sessions.

Which is cheaper?

GlitchTip, by a wide margin: free for 1,000 events, $15 for 100,000, or self-hosted. Raygun has no free plan, and Crash Reporting alone starts at $40 a month on annual billing.

Final thoughts

GlitchTip did the core job better for less: it caught every failure, put the server's code on the error page, and charged a fraction of Raygun's price, with logs and uptime included. Raygun did the job around it better: its RUM measured what shoppers felt, and its deployment page named the fix that failed.

So the decision is whether your problems live in the code or in the user's browser. If it's the code, GlitchTip covers it for the price of lunch. If it's the browser, Raygun's RUM is worth its subscription, and you can skip the APM.

Whatever you choose, don't let a monitoring agent near your error reports untested.