Autopilot
Your idea goes in. A launched product comes out.
3 followers
Your idea goes in. A launched product comes out.
3 followers
Autopilot turns Claude into a product team, not a search box. Give it a raw idea - software, a course, a service, a physical product, an internal process - and it drives you through 11 gated stages, orient to measure, one decision at a time. Under the hood: 4 commands, 4 sub-agents (plan reviewer, codebase reviewer, market researcher, release verifier), and a bundled MCP server that enforces its own rules. Claude Code needed only for software. No API keys. $15 once.
A few questions keep coming up, so here's an FAQ.
What do I actually get at the end?
A live thing people can use on its own web address, a landing page designed and written, launch assets plus a dated launch-day plan, and a first loop for tracking what happens so you know what to build next.
Claude can already brainstorm my idea. What am I paying for?
Claude answers the question in front of it. Autopilot holds the whole journey. The difference shows up on day nine, not day one: you come back after a week away and something already knows where you stopped, what you decided, what you decided against, and what the next move is.
Is it only for software?
No, and this is new since launch day. It asks what kind of thing you're making: software, content (a course, book, newsletter, template pack), a service, a physical product, or an internal tool or process. The 11 stages and their gates are identical for all five. What changes is what the middle stages produce. Claude Code is only needed for the software path; the other four run start to finish in ordinary Claude.
How does the build stage work when there's no code?
Same four beats: plan the piece, make it, prove it on something real, checkpoint it. Something real means real data for software, a real learner through a real module for a course, a real client through a real delivery for a service, a physical sample in the conditions it'll be used in, or one real team doing one real week of work for something internal. Nothing is finished because it looks finished.
What does hands-off actually mean? Am I still in control?
In the first three modes you're in the loop: work stops until you respond. In hands-off you're on the loop: work carries on without you and you watch it. Three things stay yours in every mode, that one included - the goal (what counts as done), the limits (what must never happen, what has to be asked first, what it may spend), and the verdict (whether it ships). The trade is stated when you pick rather than discovered later: you won't see every decision as it's made.
Do I have to learn anything first?
No. Type /autopilot and describe your idea in a sentence. The four commands are conveniences, not requirements. You're never asked to name a file, learn a stage number, or remember where anything lives.
Is this tested, or theory?
The build stages come from one real product taken end to end: six weeks, ten milestones, 525 tests. Being straight about the rest - that product was software and it never launched, so the launch, growth and measurement stages are written but unproven, and so are the four non-software paths. Better you hear that from me than find it out.
Will it build my product without me?
No, and be suspicious of anything claiming otherwise. It does the work and makes the drafts. You make the decisions that change direction, and you own the accounts, the money and the final call.
Full FAQ is on the listing. Happy to answer anything not covered here.