June 9, 2026·4 min read

How to Use Agent Spinnr

A five-minute setup: install the GitHub App, add an agent key, pick your triggers, and let issues resolve themselves.

Getting Agent Spinnr running takes about five minutes. Here's the whole flow.

1. Install the GitHub App

Sign in, then install the Agent Spinnr GitHub App on the repositories you want to monitor. You can scope it to a single repo while you're kicking the tires — there's no need to hand over your whole org.

2. Add a background-agent API key

Agent Spinnr doesn't run a model itself. It drives your background agent, so you bring your own key:

Cursor — grab a key from your Cursor account.
Claude Code — supported as well.

Keys are encrypted at rest. The agent acts on your account, which means the usage (and the bill) is yours and under your control.

Important: your repositories must be connected in your Cursor account for the agent to be able to access them. Connect them at cursor.com/dashboard first.

3. Choose your triggers

Decide which GitHub events should launch an agent:

Opened — fire the moment an issue is created.
Assigned — only when an issue is assigned (to a person or a label).
Labeled — e.g. an agent label as an explicit opt-in.
Edited / Commented — keep the agent in the loop as the conversation evolves.

If you're nervous, start with labeled. Nothing runs until you deliberately add the label, which makes the whole system opt-in per issue.

4. Let it run

Once configured, the loop looks like this:

An issue event fires one of your triggers.
The agent launches with the issue title and body as its prompt.
New comments and edits are streamed in as additional context.
When it's done, it can auto-open a pull request as the Cursor GitHub App bot.

You watch it all from the dashboard — live agent runs, their status, and links straight to the PRs they produce.

Tips for good results

Write issues like prompts. A clear title, the expected behavior, and a pointer to the relevant file or area go a long way.
Keep the scope small. "Fix the off-by-one in pagination" lands far more reliably than "redesign the dashboard."
Use labels as a safety valve. Trigger on a label so you stay in control of exactly what runs.
Review every PR. It's a first pass, not a final answer. Treat it like a fast junior who never sleeps.

That's the entire product. Connect a repo, point it at issues, run agents. For the why behind it, read Why Agent Spinnr exists.

2. Add a background-agent API key

Agent Spinnr doesn't run a model itself. It drives your background agent, so you bring your own key:

Cursor — grab a key from your Cursor account.

Claude Code — supported as well.

Keys are encrypted at rest. The agent acts on your account, which means the usage (and the bill) is yours and under your control.

Important: your repositories must be connected in your Cursor account for the agent to be able to access them. Connect them at cursor.com/dashboard first.

3. Choose your triggers

Decide which GitHub events should launch an agent:

Opened — fire the moment an issue is created.

Assigned — only when an issue is assigned (to a person or a label).

Labeled — e.g. an agent label as an explicit opt-in.

Edited / Commented — keep the agent in the loop as the conversation evolves.

If you're nervous, start with labeled. Nothing runs until you deliberately add the label, which makes the whole system opt-in per issue.

4. Let it run

Once configured, the loop looks like this:

An issue event fires one of your triggers.

The agent launches with the issue title and body as its prompt.

New comments and edits are streamed in as additional context.

When it's done, it can auto-open a pull request as the Cursor GitHub App bot.

You watch it all from the dashboard — live agent runs, their status, and links straight to the PRs they produce.

Tips for good results

Write issues like prompts. A clear title, the expected behavior, and a pointer to the relevant file or area go a long way.

Keep the scope small. "Fix the off-by-one in pagination" lands far more reliably than "redesign the dashboard."

Use labels as a safety valve. Trigger on a label so you stay in control of exactly what runs.

Review every PR. It's a first pass, not a final answer. Treat it like a fast junior who never sleeps.

That's the entire product. Connect a repo, point it at issues, run agents. For the why behind it, read Why Agent Spinnr exists.