Crash reporting, error tracking, logs and product analytics that run on your own Cloudflare account.
Try the live demo with six made-up apps sending real crashes, or read more on crashyard.dev.
Your apps keep the official Sentry SDKs they already use. crashyard groups their crashes into issues, turns stack traces back into readable code, and tells you when something new breaks. It is a single Cloudflare Worker with D1 and R2. Cloudflare's free plan handles a few thousand events a day, and docs/deploying.md says what fills up first.
Apps with native crashes or source maps need the Workers Paid plan. Turning their stack traces back into code takes more processing time than the free plan's 10 ms a request. crashyard finds which plan your account is on by itself, and warns on the pages that need Paid.
crashyard speaks Sentry's protocol. It isn't affiliated with Sentry.
-
Deploy it to your Cloudflare account.
-
Open it, create your owner passkey with the setup code, and invite your team.
-
Create a project for each app and give its Sentry SDK the project's DSN.
docs/deploying.md walks through every step, including deploying from a clone.
- Issues. Crashes and errors are grouped the way Sentry groups them. Resolve them now or in the next release, ignore them for a while, merge duplicates, add notes, and search with Sentry's query syntax. A resolved issue that comes back is flagged as a regression.
- Readable stack traces. Upload dSYMs, ProGuard and R8 mappings, Android native libraries and source maps with
sentry-clior the Sentry build plugins. crashyard reads them with Sentry's ownsymboliclibrary, and crashes that arrived before their files are processed again automatically. - Links to your code. Each stack frame opens its line on GitHub or GitLab, and each release lists the commits it shipped.
- Releases. See what is new, what regressed and what is gone in each release, with crash-free sessions. Crashes from App Review and TestFlight are called out.
- Logs. Search Sentry logs, follow them live, group them and jump to the issues in the same trace.
- Product analytics. Mark a Sentry log as an action and crashyard counts it, with no second SDK. Daily, weekly and monthly active users, new users and stickiness; insights by platform, OS, country, release or a property; funnels you can save, with conversion by release; weekly retention; flows; each person's timeline; a Product dashboard template; and alerts on an action's count. Actions and errors share their users, so a funnel shows how many people stopped at a step after an error, and retention compares those who hit one with those who didn't. Actions count towards the daily log limit and are kept 365 days. docs/linking.md shows how to send them.
- Dashboards. Build charts, numbers and tables from issues, events, logs and actions, or start from the Crashes or Product template.
- Monitors. Cron monitors for scheduled jobs and uptime checks for URLs, with public status pages showing the checks you choose from any of your projects. Post incidents, notices and planned maintenance with their updates and the services they affect, with your own look and your own domain.
- Cloudflare Workers. Catch the runs that run out of CPU time or memory, which a Worker's SDK never gets to send. On Workers Paid, for Workers on the same account.
- Performance. Transactions by p95, how each release changed them, and the slowest ones with their spans.
- Notifications. Alerts through Telegram, Slack, Discord, email or a JSON webhook, routed by alert type, project, level and environment, with quiet hours and a weekly summary. Resolve, ignore or assign an issue right from its alert: in the chat on Telegram, in one tap on crashyard from the others. Every alert sent in the last 30 days is listed, with where it went.
- Moving from Sentry. Bring a project's issues over with their latest event, counts, status and comments. Events from the SDK keep coming to the same issues. Until you switch, every event can go on to Sentry too.
- MCP server. Lets an AI agent read issues, events, logs, traces, releases, performance and monitors, triage issues and post to status pages. See docs/mcp.md.
- People and sign-in. Passkeys, invites and four roles, from owner to viewer, with recovery codes and an optional Sign in with GitHub.
| SDK | Status | Stack traces |
|---|---|---|
| Apple, sentry-cocoa | Works | Symbolicated from dSYMs |
| Android, sentry-android with NDK | Expected to work | From the R8 mapping and native libraries |
| JavaScript in the browser | Works | From source maps |
| Node.js | Works | As sent |
| Python, Ruby, PHP, Go, Java, .NET | Works | As sent |
| Elixir, Dart | Works | As sent |
| Rust | Works | As sent, or from debug files |
| React Native, Flutter, Unity | Events arrive | Raw frames, mappings not read yet |
To connect an app, see docs/linking.md. For exactly which parts of Sentry's protocol and API work, see docs/sentry-compatibility.md.
crashyard sends no telemetry. The only request it makes on its own is a daily check with GitHub for a newer release, so the overview can tell you about it. It stores no IP addresses, keeps a user only as a hash of the app's user ID, and removes anything that looks like a password, token or email address from events and logs.
pnpm install
pnpm wrangler d1 migrations apply DB --local # creates the local database's tables
pnpm dev # the dashboard at http://localhost:3000, with local D1 and R2
pnpm seed # fills it with a month of sample crashes, logs and transactions
pnpm check # lint, formatting, types, unused code, tests and the build, as CI doescrashyard is released under the MIT license.