How can we help?

PromptPilot turns a rough app idea into a build-ready blueprint — a multi-file set of tool-ready files for AI coding tools. It reads your idea, asks adaptive questions, and flags the decisions still open. It's a single Studio: describe your idea, answer the questions, then refine, version, trace, and export the files.

No account is needed to start — PromptPilot is local-first, and you never bring your own API key. The AI-powered steps (planning, generating, refining) ask you to sign in first so usage limits stay fair. Signed out, projects stay in this browser; signed in, they also sync to your account so they work across devices.

It's only a wording choice. Both modes ask the same questions and produce the same blueprint — Simple uses plain language, Developer uses technical terms (data, auth, APIs). You can switch anytime in Settings.

After you describe your idea, PromptPilot builds an App profile (what kind of app it is, who it's for, what data it keeps) and asks questions tailored to it — only what's actually missing for your app. Which questions show is decided by a local, deterministic engine, not AI, so the same answers always lead to the same path.

Recommend fills in a sensible default for that question so you can keep moving. Other lets you type your own answer — if it's ambiguous, PromptPilot may ask a couple of quick follow-ups to sharpen it.

As you answer, PromptPilot tracks two lists — locked decisions (what you've settled) and missing decisions (what's still underspecified or conflicting, grouped by urgency). For any missing one you can answer it, use a suggested default, or mark it for later, all in one click. After generation the Workflow view keeps showing how the settled decisions connect.

Before generating, PromptPilot shows an editable product description — a plain-language summary of exactly what your V1 is, written from your answers (features, storage, sign-in, payments, look & feel, stack). Edit anything that's off — by typing or voice — flip to the UI preview to see a mock in your chosen colours and style, then Generate Blueprint. That description is the source of truth the files are built from.

A spec plus tool-ready files — universal files like PROJECT.md, BUILD_PROMPT.md, FEATURES.md, and MVP_SCOPE.md, plus a single agent-instruction file. Pick the AI tool you build with at Export — Codex, Cursor, Claude Code, GitHub Copilot, and more — and the instruction file is renamed to that tool's convention. Adaptive files (auth, payments, database schema) are added only when your app needs them.

Two ways. Add details opens a panel where you can type (or dictate) a note and optionally attach a reference image — the note can name several things at once and across files, and PromptPilot works out which files each part belongs to and shows you the planned edits to confirm before anything is applied (if the note is vague, it asks a few quick questions first). Leave the note blank and it simply adds detail to the file you have open. Or use the pen button on the open file to edit it by hand — that's free and costs no credits. Every change is reversible with Undo / Redo.

PromptPilot generates your full blueprint if you don't have one yet (normal cost — starting a build adds nothing extra), freezes an exact copy as your build baseline, and prepares the package for your coding tool: a ZIP arranged for the tool, a copy-paste bundle for a chat tool, or the plain package. PromptPilot doesn't write the code — your coding tool does. The frozen version becomes read-only so you always know exactly what you built from; when you want something different, ask in the change box — small tweaks apply to your working copy, bigger ones become the next version. Because freezing locks the version, it's part of the paid plans — on the free plan you still get the whole blueprint and can download it from Export, unlocked.

Press the ⊕ in Studio and describe what's next. PromptPilot plans a focused increment from your existing plan and generates a new version's files — your earlier versions stay intact.

Roadmap maps what your app is made of (Frontend / Backend / Security / Deploy); Specs describes what it is (Features / Decisions / Stack). They're prepared when you first open them and refreshed when you switch versions — file edits don't spend credits rebuilding them. Each panel has an (i) explaining its legend.

Click the workflow symbol in Studio's toolbar (just left of Copy) — or open the Workflow pane at the review checkpoint. It shows how your app works, end to end — who starts each flow, which screens they see, what the server does, where data is saved, what protects it, and which outside services are involved — as a compact flow you can expand, search, and filter. It's labelled Planned from your blueprint: it describes the app you're designing, honestly, never claiming code exists. Built locally from your answers — no AI, no credits — and part of the paid plans (free users see an example demo).

Open the Export tab in Studio's right panel: Copy all files, Download (ZIP), or Download JSON (the whole project), and copy the ready-to-paste build prompt. The pack ships tool-agnostic — the agent-instruction file is AGENTS.md, which every major coding tool reads. A share link is offered when cloud storage is configured.

Open Settings → MCP. The Setup Center generates ready-to-copy configuration for Claude Code, Codex, VS Code / Copilot, Cursor and Gemini CLI — trusted coding connectors (GitHub, Supabase, Playwright, …) and models you run yourself (Ollama, LM Studio, llama.cpp, vLLM, or any compatible endpoint). Everything is generated locally, with no AI and no credits, and PromptPilot never installs software or edits your tool's settings from the browser — you copy the setup into your own terminal or config file. Local-model prompts go straight from your machine to your endpoint; PromptPilot never proxies or meters them.

Signed out: on your device, in your browser's local storage — projects, answers, generated packs, version history, and preferences. Signed in: projects are also synced to your account (stored with our hosting provider) so they work across devices; the local copy remains, so nothing depends on the network.

Settings → Delete all data wipes every project, session, and preference from this browser (you'll confirm by typing DELETE); clearing your browser's site data has the same effect. Projects synced to your account aren't touched by that — delete them individually, or use Settings → Delete account to erase your account and everything in it. All of these are permanent.

Only the text you provide when you trigger an AI action (plan, generate, refine, improve), sent to the configured provider solely to produce that result — review that provider's own privacy terms. With no AI key configured, processing is fully local with no external calls.

Free covers one full blueprint a month — enough to take an idea all the way to V1 — plus 5 voice notes a month (90 seconds each). Paid plans add editing & refining, voice input with no monthly cap, versions (V2, V3…), the Workflow view, Premium output, and a much bigger monthly credit pool. See the Pricing page for current plans.

The final deliverable (your generated blueprint) can render on a more capable model when your plan allows it — otherwise it renders Standard. Premium output costs more of your monthly credits. The app only ever shows Standard vs Premium, never provider names.

It shows how much of your allowance is left. On a paid plan it leads with your weekly cap (with a 5-hour burst window underneath); free shows the monthly pool. Every AI action spends credits — generating a blueprint spends the most; edits and versions spend less. The weekly cap refills each week, the pool each month.