AI Automation

Let Claude diagnose feedback issues — and open fix pull requests — automatically

Choose your automation level

Every DebugGo project picks what happens after feedback becomes a GitHub issue. Change it anytime from your project's settings — no repository changes needed.

Create issues only

Feedback becomes a labeled GitHub issue. No AI runs.

AI diagnosis

Claude analyzes each issue, posts a root-cause report, and retitles it.

AI fix & PR

Diagnosis, then Claude implements the fix and opens a pull request.

It works through labels: DebugGo tags each issue it creates with debuggo:diagnose or debuggo:autofix, and one static workflow file in your repository reacts to them. Switching levels in the dashboard changes the label on future issues — the workflow file never changes. You can also add either label to any issue by hand to run the AI on demand.

Runs in your repo, on your account

The AI runs as a GitHub Actions workflow in your repository, authenticated with your own Anthropic credential — a Claude Pro/Max subscription token or an API key. DebugGo never holds your key and never bills you for tokens.

Setup

Step 1: Add the workflow file

Fastest path — install the setup skill and let Claude Code do everything (workflow file, secret, and PR) with one command:

Then run /debuggo-setup in Claude Code inside your repository. It also checks and repairs existing installs.

Prefer doing it yourself? Fetch the workflow directly and commit it:

The file lives at .github/workflows/debuggo.yml and is identical for every repository — don't edit it. (You can also copy it from your project's GitHub settings in the dashboard.)

Step 2: Add ONE repository secret

With a Claude Pro or Max subscription (recommended — runs are covered by your plan):

Or with an Anthropic API key from console.anthropic.com (metered billing):

Step 3: Install the Claude GitHub App (recommended)

Install github.com/apps/claude on your repository. The workflow then pushes fixes and opens PRs as that app — no repo settings needed, and your CI runs on fix PRs. (DebugGo's own app never needs write access to your code.)

Skipping it? Autofix falls back to the workflow's built-in token, which additionally requires enabling Allow GitHub Actions to create and approve pull requests in Settings → Actions → General, or from the CLI:

Step 4: Pick the level in your dashboard

In DebugGo, open your project → Settings → GitHub. Connect the repository, enable Auto-create Issues, choose AI diagnosis or AI fix & PR under AI Automation, and save.

That's it

Every new feedback submission now flows: widget → issue → diagnosis → (optionally) fix PR — with no one touching the keyboard.

Verify it works

No need to wait for real feedback — add the label to any open issue and watch the DebugGo Automation run appear in your repo's Actions tab:

A successful run posts a 🔍 DebugGo Diagnosis comment and gives the issue a descriptive title.

Troubleshooting

Run fails with a 401 authentication error

The secret is missing, mistyped, or expired. Subscription tokens expire — refresh with claude setup-token and set the secret again.

No workflow run when an issue is created

The issue has no debuggo:* label — the automation level isn't set in the dashboard, or the connected repository doesn't match this one.

The fix job shows as skipped

Expected when the label is only debuggo:diagnose — the fix job runs only for debuggo:autofix.

AI automation builds on the GitHub integration — issues must be auto-created by DebugGo for the labels to be applied. Codex support is on the roadmap.