Skip to main content
Autter Skills are Agent Skills that teach a coding agent — Claude Code, Cursor, Codex, or anything else that supports the standard — how to install Autter Runtime into a codebase. Point your agent at them once, and it inventories your repo, picks the right integration per service, and wires up telemetry without you writing the instrumentation by hand.
This is the fastest path to a working setup. If you’d rather do it yourself, see Quickstart (npm packages) or Using Autter Runtime without npm (any other language).

Install the skills

Requires only Node/npx on the machine running the agent — the target codebase doesn’t need Node at all.
This drops the skill files into your project (or wherever your agent looks for skills). Then just tell your agent:
Use the skills to install Autter Runtime in this project.
Want a single skill instead of the full set — for example if you already know your stack is Python — install just that one:

What happens when your agent runs it

  1. autter-runtime-setup runs first. It inventories every service in your repo (languages, frameworks, monorepo packages) and asks you for an ingest key if you haven’t already created one — see Getting an ingest key below.
  2. Each service gets routed to the right style skill automatically — you never have to pick one yourself:
  3. The agent verifies telemetry arrives. In production, it inspects existing traffic and exporter failures. Controlled test errors and requests belong only in a local or isolated test environment.
  4. Logging setup checks capability first. The agent checks installed Node/Next.js exports, deployed /v1/logs ingestion, and platform evidence support. It reports pending releases or deployments instead of adding unsupported APIs. Other languages need a logs exporter and logger bridge; browser/edge capture keeps its own APIs.
  5. Customer operations get meaningful context. Selected checkout/job paths record nested context, measured steps and explicit business outcomes while preserving existing loggers. The agent checks lifecycle flushing, stored logs and trace-linked failure evidence separately, then records the supported instrumentation convention in existing contributor instruction files.

Set up operation logging

After installing the updated skills, ask your agent:
Update this project’s Runtime setup for structured logs and operation evidence. Check release and deployed support, preserve our existing logger, instrument checkout and background jobs, and verify stored logs and linked failure evidence.
The skills use runtimeLogger for diagnostics and explicit failed operation outcomes or exception capture for issues. A returned callback defaults to success; a payment rejection without a throw needs a declared outcome. The APIs require Node/Next.js SDK and ingester 1.4.0+ (compatibility table), followed by an application redeploy; updated skills alone do not provide that release. See Operation logging for the APIs, Node/Next.js examples, OTLP logging in other languages, delivery limits, and verification.

Set up endpoint regressions

Use the setup skills to inspect existing SDK initialization, update compatible packages, and configure request histograms and release identity. Node/Next.js examples opt into slow successful trace retention. External OTel exporters need explicit-bucket delta histograms. Platform access does not change your deployed applications. Reuse their existing providers and deploy the required SDK configuration. Self-hosted ingesters need the versions in the compatibility table. See Endpoint regressions for the complete setup, team feedback, and human review of draft fixes.

Connect external logs

Ask autter-runtime-setup to connect your existing Sentry, PostHog, Grafana/Loki, Datadog or webhook source for the selected repository. The agent uses repository Runtime settings and checks stored records, analysis state and the actual draft PR URL separately. Connector-only setup does not require an ingest key or application SDK install. Read External log sources for provider credentials and automation policy.

Getting an ingest key (application instrumentation)

Create one on your Autter dashboard: Settings → Access Tokens → Runtime ingest keys → Create key. Pick the repository, then choose:
  • Server (autter_rt_…) — secret, for backends and relays. Keep it in env vars, never commit it.
  • Client (autter_rtc_…) — publishable, for browser-only apps with no backend. Scoped to the exact origins you register it for.
Your agent will ask for whichever key kind the service it’s instrumenting needs and the environment variable name. Do not paste the key value into chat. Runtime ingest keys are separate from CLI/PAT and MCP access tokens. Server keys send OTLP logs; browser client keys cannot use /v1/logs.

Updating or removing

Re-run the install command to pick up new skill versions:
To stop using a skill, delete its folder from wherever npx skills add placed it in your project — skills are just files, there’s nothing to uninstall from a package manager.

Quickstart

The manual npm-package path, if you’d rather set it up by hand.

Without npm

Wire up any language with a plain OTel SDK or OTLP/HTTP.