SWEny

SWEny is a platform for building AI-powered engineering workflows. Define what Claude should do at each step, give it the right tools (skills), and let the DAG executor handle the rest. Triage is the first built-in workflow — it monitors your production alerts, investigates errors, and opens fix PRs automatically.

Get Started in 5 Minutes View on GitHub

What it does

SWEny workflows are DAGs of AI-powered nodes. Each node gives Claude an instruction and a set of skills (tools). Claude does the work — queries logs, searches code, creates issues, opens PRs — and the executor routes to the next node based on natural-language edge conditions.

Gather context — Claude uses observability and code search skills to understand what happened
Take action — Claude creates issues, writes fixes, and opens PRs using the tools available at each node
Report results — Claude sends summaries through Slack, Discord, Teams, email, or webhooks

Triage is the first built-in workflow. It runs on a schedule, finds novel production errors, and fixes them without waking anyone up. See How Workflows Work to learn how the executor works and how to build your own workflows.

See it in action

Want to understand the full flow? Read the end-to-end walkthrough — a real error spike from Datadog through to a merged PR.

Skills — connect to your stack

GitHub

Search code, read files, create issues, open pull requests, list commits. The most-used skill across all workflows.

Observability

Sentry, Datadog, and BetterStack skills — query errors, search logs, list monitors, and pull metrics from your existing monitoring stack.

Issue Tracking

Linear and GitHub Issues — create tickets, search for duplicates, update status, and link PRs automatically.

Notification

Slack, Discord, Teams, Email, Webhook — send results wherever your team works.

Three entry points

All entry points are powered by the same DAG executor, which manages skill resolution, node orchestration, and edge routing.

GitHub Action — runs on a cron schedule in your CI. No servers, no infrastructure. This is the primary deployment for production monitoring.

CLI — run sweny triage from your terminal. Ideal for testing, debugging, and local development. Supports dry-run, streaming output, and all the same providers as the Action.

Studio — visual DAG editor and live execution monitor. Design workflows by dragging nodes, simulate runs, and watch live executions in real time.

Quick install

- uses: swenyai/sweny@v4
  with:
    claude-oauth-token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
    dd-api-key: ${{ secrets.DD_API_KEY }}
    dd-app-key: ${{ secrets.DD_APP_KEY }}
    dry-run: true # Remove when ready to create PRs

3 secrets. GitHub Issues by default. Full setup guide →

27dc3231df485f4bde7b8aed32f4844ebe493455