Getting started

Getting a product onto self-driving mode takes three connections and a few minutes. This page walks the whole path: create the product, run the Setup Wizard, then turn on the loops.

1. Create a product

Everything in 1mn is scoped to a product — one SaaS you're running. From the top bar choose + New product, give it a name and your website URL.

When you add a website, an onboarding run kicks off automatically: it reads your site and drafts a business profile, a brand voice guide, and a first pass at your competitors — so the agent starts with context instead of a blank page. See How the agent learns your business for what each artifact contains and how loops reuse it.

2. Run the Setup Wizard

The Setup Wizard (/setup) is the guided path to self-driving mode. It walks three connections in order and derives each step from server state, so you can refresh or come back later and it resumes exactly where you left off.

Step 1 · Connect Claude Code

Coding and docs work runs on Claude. Connecting your own Claude Code means those runs use your Claude Pro/Max subscription instead of platform credits. In a terminal on your machine (with Claude Code installed), run:

claude setup-token

Paste the token it prints (it starts with sk-ant-oat01-…) into the wizard and click Connect.

Step 2 · Connect GitHub

GitHub is required for any loop that writes code or docs. Click Connect GitHub → to install the 1mn GitHub App on the account that owns your product's repository — you pick exactly which repositories it can see. You'll be sent to github.com and returned to the wizard. If the app is already installed on an account, the wizard shows a Use this → shortcut so you can reuse it in one click.

Then choose the repository the agent should work in. This becomes your product's selected repo — the only one 1mn clones, edits, or opens PRs against.

Step 3 · Switch your workspace on

The agent is about to do real work in your repo, and agent runs are subscription-gated, so the wizard asks you to subscribe before it spawns any work. 1mn is a flat $49/month per workspace — unlimited runs, no metering. Click Subscribe → to complete checkout, then the wizard continues on its own. See Billing for the full model (and the separate prepaid ad budget).

Step 4 · Install the SDK

With the connections in place, the agent installs the 1mn SDK for you — it opens a pull request that drops the snippet into your repo and wires identify() / reset() into your auth pages. It verifies the build before opening the PR.

Review the diff right in the wizard and click Merge & deploy (or open it on GitHub first). Your own CI deploys the merge. Setup completes the moment your deployed site sends its first event — that's when the dashboard flips to live.

That one SDK install is all the signal the loops need — pageviews, grouped + symbolicated errors, web vitals, in-app feedback, and (once you turn it on) session replay all flow from it.

3. Talk to the dispatcher

The left sidebar is your conversation with the primary agent. It doesn't do the heavy lifting itself — it routes. Ask in plain language ("audit my SEO", "draft a launch ad", "why are users hitting this error") and it spawns the right background task.

4. Turn on the loops

Open the Loops tab. Each loop is a skill the autonomous engine wakes on a schedule — business digests, SEO audits, content, monitors, and more. Every card shows its cadence and next wake; open one to set its schedule or change how often it runs. From then on it wakes on its own, decides whether there's anything worth doing, and hands back to you only when it needs a decision.

See the loops for the full catalog and what each one does.

5. Review and approve

Loops surface their work as tickets and drafts. Low-risk, reversible work applies on its own; anything irreversible — a production deploy, money out, a live ad — pauses for your approval. You review on the Tasks board, or right in chat. See Tickets & approvals for how the gate works.

Next steps