Node.js 
BunError
The hardest production bugs disappear before you can inspect them.
Logs, traces, and stack traces show you where something failed. But when the answer depends on runtime state you never recorded, debugging becomes a reconstruction exercise.
Typical production debugging
Production error
Read the stack trace
Search logs and traces
The variable you need was never recorded
Add logs
Redeploy
Wait for it to happen again
Add a probe
Wait for it to happen again
Execution recorder
Follow the values back to the cause.
Rasputin records function calls with their arguments and return values, so you can step through a production failure yourself.
POST /api/shipping/quote 500 16ms
buildShippingQuote threw · 1ms src/shipping/build-shipping-quote.ts
Loading source…
Current call
buildShippingQuote 1ms
Arguments
cartTotal: 84.5
rate: undefined
Threw
TypeError
On-demand recording
Get the missing context without another deploy.
Start a capture session for a route or job, and Rasputin records its upcoming runs, including the successful ones.
Executions production
POST /api/checkout
HTTP requestCapturing 0/5
12m left
Recordings
| Outcome | Kind | Status | Duration | When |
|---|
Investigation agent
Turn runtime evidence into a reviewable fix.
Rasputin investigates the failing release in an isolated sandbox, using the recording and your connected tools. It can run checks, develop a fix, and prepare a change for review.
Runtime
Source
Deploy
Rasputin
Investigation
Root cause
Fix generated
Less noise. More context.
Fewer duplicate alerts.
Semantic grouping recognizes related failures from their code path and error context, even when the messages differ.
Focus on application failures.
Scanner traffic, browser-extension errors, and one-off network blips are marked as noise and kept out of your alerts.
Context from your stack.
Connected tools let investigations check your logs, metrics, issues, and runbooks.
Connect, install, and deploy.
Connect GitHub and an alert channel, install the SDK, and deploy as usual.
GitHub connected
Rasputin-AI/rasputin
Slack connected
#production-alerts
Discord connected
#production-alerts
Questions
Setup guides and configuration are in the documentation.
Which runtimes are supported?
Long-running Node.js and Bun servers, including Elysia and SvelteKit apps. Serverless and edge runtimes aren't supported. Some build setups need an extra instrumentation step. Check compatibility
Can I use recordings without the investigation agent?
Yes. Every recording opens in the replay debugger, and investigations only run when you start one.
What gets recorded, and what is redacted?
Arguments, return values, and call history, within size limits. Common secrets, such as passwords and authorization headers, are redacted at ingest, after the data leaves your app. Set scrubPII to also redact personal data, or exclude files you never want recorded. Redaction
How is Rasputin billed?
Pro is $30 a month with $15 in usage credits, and Pro+ is $90 with $60. Error capture, grouping, and recordings are included. Investigations and follow-up questions use credits. Pricing