1mn.ai docs

FAQ

Quick answers to the questions founders ask first — what a loop is, what runs on its own vs. waits for you, how billing works, and what 1mn does with your code.

What exactly is a "loop"?

A loop is a system that, on a schedule or trigger, discovers work, hands it to an agent, verifies the result, records what happened, and decides the next move — handing back to you only at the edges. It's not a one-shot prompt: it acts, observes, and repeats, which is what lets it recover from failure and make progress across many runs. See How loops work.

What runs on its own, and what waits for me?

Reversible, low-risk work applies on its own. Anything irreversible pauses at awaiting review with the artifact attached. A few actions always require your explicit approval:

Always gated

Setting an ad campaign live, spending from your budget wallet, and anything touching auth, billing, or database migrations.

Each loop's mode — draft-only or recommend — is fixed by its blast radius, and the irreversible steps above always stay a manual gate you approve; there's no autonomy dial to loosen. You enable each loop and set its schedule from the Loops view, the Rangers tab, or the loop's own tab. See How loops work and Tickets & approvals.

Does 1mn store my code or a GitHub token?

No long-lived token. When you connect GitHub, 1mn mints short-lived, installation-scoped tokens on demand, and code loops clone your repo into an isolated sandbox for the duration of a run. The deliverable comes back as a pull request you review and merge.

How much does it cost?

$20/month per workspace, billed through Polar, with 10 task credits included every cycle (1 credit = 1 task — a ticket's first agent run). The signal layer — session replays, users, feedback — is free without any subscription, and chat, scheduled routines, and retries never consume credits. Connect your own Claude subscription and tasks become unlimited. See Billing & subscription.

Is ad spend part of my subscription?

No — ad spend is separate. You top up a prepaid budget wallet, which funds your Meta and Google ad accounts and caps total spend. Campaigns pause automatically when the wallet hits zero, so a loop can't overspend.

Can I use my own Claude subscription?

Yes. Connect your Claude Pro/Max subscription from Integrations and task runs bill to your Claude account instead of the platform key — they consume zero workspace credits, making tasks unlimited. It's optional — the primary chat always runs on the platform key.

Which integrations do I actually need?

Only GitHub is required, and only for loops that write code or docs. Everything else is optional and each loop degrades gracefully without it: Revenue (Polar) feeds your daily brief, Search Console powers SEO, and Telegram lets you reply to the agent from your phone.

Will Cast or the Reddit loop touch production?

No. Dogfooding Cast personas drive a real browser, so you aim them at a staging URL with a throwaway account — never production. The Reddit loop is read-only and draft-only: it surfaces threads and pre-drafts a comment, but never posts — you post manually.

Where do I review everything a loop produces?

On the Tasks board. Every piece of work — whether you asked for it or a loop generated it — lives as a ticket you can watch, comment on, and approve. Commenting on a ticket re-engages the agent with full context.

On this page