Skip to main content
Autter Runtime tracks runtime errors and usage from your frontend and backend. It deliberately does not ship the full OpenTelemetry browser SDK to your users:
  • the browser gets a dependency-free, <5 KB error and usage tracker
  • your server keeps real OpenTelemetry
  • Autter Runtime’s OTLP ingester receives both and writes them to ClickHouse in a compact, per-repository data model
Fastest way to get set up — let a coding agent do it:
Then tell your agent: “use the skills to install Autter Runtime in this project.” See AI agent skills for details.

Quickstart

Go from zero to seeing data in ClickHouse in a few minutes.

Installation

Install the package for your stack — browser, Node, or Next.js.

​
Find endpoint regressions

Autter can detect slower HTTP endpoints from request-duration histograms, compare normal and slow traces, collect team feedback, and generate draft fix pull requests when evidence supports a change. Your team reviews, merges, and deploys fixes. Automatic rollback is not part of this pipeline. Your service must send compatible metrics and retain useful traces. Platform access does not update your SDK settings. See Endpoint regressions for setup and data requirements.

​
Packages

A runnable demo lives at examples/express-app — browser tracker → relay → ingester, and the OpenTelemetry server tracker, against a Docker Compose ClickHouse.

​
Supported stacks

Errors group the same way for every stack: the ingester parses each language’s native stack format, so the same defect stays one issue across re-deploys and unrelated errors that share a message stay separate. See Architecture & data model for how grouping works. See Stack integrations for per-stack snippets, or Using Autter Runtime without npm if your service has no @autter/* package at all.

​
Keys: frontend vs. backend

Two credential types keep the frontend and backend cleanly separated:
When your app has a backend, prefer the relay: the browser posts to your own server, which forwards with the server key. No key ever reaches the browser, and ad-blockers can’t tell it apart from your own API traffic.

​
Design principles

  • Errors are 100%, everything else is sampled or aggregated. Raw error occurrences are always kept (14-day TTL); successful traces are expected to be sampled upstream (0.5–1%); usage is stored as 1-minute rollups (90 days).
  • Per-repository analysis. Every row is keyed by org_id + repository_id.
  • Privacy by construction. No cookies, no DOM, no request/response bodies, no emails, no full URLs with query strings.
  • OTLP-compatible at the ingestion layer, not inside a 3 KB browser script.
Autter Runtime is open source under the MIT license: Autter-dev/autter-runtime.