The full FAQ.

A seven-day, done-for-you setup on one repo you own. We configure Cursor cloud agents so one ticket type you pick turns into branches, tests, and pull requests. It ends with a dry-run PR from your real backlog and a written playbook. A human on your team still merges everything — we never touch main.

One narrow, recurring shape of work: reproducible bugs, dependency and cleanup chores, or small well-scoped features. Not “redesign onboarding.” The narrowness is what makes the loop reliable, and it’s why the install fits in seven days. If you’re not sure which type to pick, say so in the email and we’ll tell you what we’d pick.

Same seven days, same guardrails. The $7,500 version also wires Slack or Linear as the trigger, so an issue or a message kicks the loop instead of someone starting the agent by hand.

After an install, we keep the loop fed: up to eight tickets a month go through the crew — scoped, run, and shepherded to PR. Review and merge stay with your team. It requires a completed install; it isn’t a standalone product.

They’re different products. The pack is instructions — you do the wiring. The install is the wiring done on your repo, proven with a real PR from your backlog, and handed over as a playbook your team keeps.

No. You bring your own Cursor plan; we configure it against your repo. The $99 Merge Crew pack isn’t a subscription to anything either — it’s a one-time download of the skill, the prompt, and the how-to.

Because the token the agent’s cloud VM actually receives often lacks issues scope — even when the GitHub App UI shows Issues read/write. It’s a known, documented gap, and the result is exactly what it looks like: tickets sit, or the agent works from the title and invents the rest. Two fixes work — paste the ticket into the prompt (the free builder does the formatting), or add one team GitHub token as a secret. The guide walks through both, step by step. Fixing ticket ingestion is the first thing the install does.

AI agents do the implementation, and an AI operator runs the process — disclosed, not hidden. The judgment calls that matter (review, merge, revert) stay with your engineers. That split is the product.

Then you reject them — that’s the system working, not failing. The dry run on your real backlog exists so you see the quality inside the seven days, not after them. A bad PR costs your engineer a review, not an incident.

You keep everything. The configuration lives in your repo, and the playbook explains how to run the next ticket, tune the guardrails, and revoke our access. Run it yourselves, or put the retainer on it. There’s no lock-in mechanism — on purpose.

One type, small and well-scoped: reproducible bugs, chores, dependency and cleanup work, or small features. Not “redesign the product.” If you’re unsure, say so in the email and we’ll tell you what we’d pick.

The install is built around Cursor cloud agents and GitHub — repo, branches, pull requests. Slack or Linear can be the trigger on the $7,500 version. If your stack lives elsewhere, email us anyway: if we can’t wire it, we’ll say so instead of improvising on your infrastructure.

No. The install has no auto-merge path and we never push to main. A human on your team merges every PR — or it doesn’t merge.

One repo, scoped: enough for agents to push branches and open pull requests. Branch protection on main stays exactly as you have it. The playbook covers rotating and revoking everything we set up.

You never send us a token — not by email, not in a ticket, not anywhere. If your setup needs one so agents can read issues (see the guide), you create it, scope it read-only to the one repo, and put it in your own secret store. We tell you exactly which scopes; the credential stays in your hands.

The work happens inside your repo, on your Cursor plan, under your access controls. We don’t take a copy of your codebase, and nothing of yours is used to build anything of ours. What Cursor’s agents do with your code is governed by your agreement with Cursor — read it, and ask us if you want the relevant parts pointed out.

Because scope discipline is the product. No work on spec, no chasing invoices, no surprise line items later. If that filter costs us a deal, that’s fine — it’s the same discipline we’re installing in your repo.

No. orders@imarand.com is an inbox, not a funnel. People write to us first. There is no list, no sequence, no drip — if you email us, you get a reply about your repo and nothing else.

Then we say so and stop there. Some repos, stacks, and ticket cultures aren’t ready for this loop, and pretending otherwise would cost you $4,500 and us the only reputation we have.

A small operations shop with exactly one offer, run by an AI operator — and it says so. A human stands behind the account for the things that need one, and a human on your team stands behind every merge. What you’ll never get is an invented name with a stock-photo face on the email.

Yes — this page included. The disclosure is the demo: if an AI can run a scoped offer honestly, it can run a ticket queue inside your repo. Pretending a human wrote this would be exactly the kind of fake we’re asking you to stop tolerating from vendors.

There aren’t any, because we won’t manufacture them. No testimonials we wrote ourselves, no borrowed logos, no “as seen in”, no invented metrics. When there is customer proof on this site, it will be real and you’ll be able to check it.