Quickstart

From signup to the first diagnosis on your own code takes less than 10 minutes, and you never write YAML.

Before you start

• A GitHub repository on which you can install apps.
• For your runner: an Anthropic, OpenAI or OpenRouter API key, a Bedrock or Vertex account, or a ChatGPT plan for Codex. For MCP mode: the agent you already use. For Managed: nothing.

1. Create the account

Sign up with email or GitHub. Choose software company for your own product, or agency if you support apps of clients.

2. Connect the repository

Install the GitHub App on the repository. Loopback opens a pull request named Add Loopback with .loopback.yml, the settings you chose, and, when jobs run in your GitHub Actions, the workflow that starts them. Merge it; there is nothing else to write.

# .github/workflows/loopback.ymlname: Loopbackrun-name: Loopback ${{ github.event.client_payload.job_id }}on:  repository_dispatch:    types: [loopback-job]permissions:  contents: write  pull-requests: write  id-token: write          # OIDC sign-in to Loopbackconcurrency:  group: loopback-${{ github.event.client_payload.job_id }}jobs:  run:    runs-on: ubuntu-latest    timeout-minutes: 75    steps:      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1        with:          fetch-depth: 0          persist-credentials: false      - uses: actions/cache/restore@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0        if: ${{ github.event.repository.private }}        with:          path: .vexp          key: loopback-index-${{ github.run_id }}-${{ github.run_attempt }}          restore-keys: loopback-index-      - name: Install bubblewrap        run: sudo apt-get update -q && sudo apt-get install -y bubblewrap      - uses: loopback-dev/action@v1        with:          agent-key: ${{ secrets.LOOPBACK_AGENT_KEY }}          api-url: "https://api.loopback.so"      - uses: actions/cache/save@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6.1.0        if: ${{ !cancelled() && github.event.repository.private }}        with:          path: .vexp          key: loopback-index-${{ github.run_id }}-${{ github.run_attempt }}

# .loopback.yml — everything is optional, defaults are auto-detectedagent:  mode: runner  name: claude-code  auth: api-keyautonomy: approvetest: npm testrelease:  tag: semver  merge: squash  changelog: trueshare: snippets

Every key of .loopback.yml is optional: see the configuration reference.

3. Choose where the AI runs

The setup asks one question: which agent you use and how you pay for it. The answer picks the execution mode and the credentials, and blocks the combinations the providers forbid. See execution modes.

4. Connect your support channels

• The inbox: every workspace gets an address on the inbound domain right away. Forward your support address to it, and verify your domain to send replies from it.
• The report widget: one script tag in your app, see the widget.
• Logs, optional: Sentry, Datadog, CloudWatch or an endpoint, see connectors.

5. Watch the demo ticket

Right after the repository is connected, Loopback runs a demo ticket on your code at its own expense, once per organization. It indexes the repository, picks the most frequent real error, an open issue, or a plausible defect written as a customer email, and shows the email, the status page, the diagnosis with files and lines, the reply draft and a draft pull request marked as a demo. The demo runs in our Managed sandbox whatever mode you choose, and its clone is deleted at the end of the job.

Next steps

• Invite your team: owners and admins see everything, reviewers approve, client viewers only read.
• Set the autonomy level: Copilot, Approve or Autopilot by category. Tags and releases always wait for a person.
• Read how it works: the states of a ticket and the rules no setting changes.