Documentation

Download the app

doable runs on your own computer, so it can build, run, and preview real apps locally before anything goes live. It's a free download for Mac, Windows, and Linux.

Prefer the dashboard? You can sign in and manage projects, billing, and domains at doable.do in any browser. The app and the dashboard share the same account.

Connect your AI

doable doesn't sell you tokens. It uses the AI subscription you already have. On first launch, pick the agent you want to build with and sign in with your existing plan.

You can switch agents per project from the model picker at the bottom of the chat. doable keeps the connection on your machine, so your AI subscription is used directly, not proxied through us.

Build something

Open the app, start a project, and describe the tool you want in plain language. The agent builds it for real, in a workspace on your computer, while you watch.

Preview

The Preview pane shows your app running live as it's built. doable starts the dev server for you and reloads on changes, so you see the real thing, not a mockup.

• Web apps render in a framed browser at their local URL, with desktop / tablet / phone sizes.
• Mobile apps preview in a phone frame; native-only apps offer an “Open in Simulator” option.
• Desktop apps show a live mirror of the running window.

Publish

When a tool is ready, hit Publish. doable puts it where people can actually use it, and the right destination depends on what you built:

A project can have more than one of these at once (a web app plus an API plus a scheduled job, for example). doable treats each as its own service and publishes them together.

Share with your team

A project isn't only yours. Open the Share sheet, invite people by email, and they get the same project in their own copy of doable. There are no branches to merge and nothing for a non-developer to learn.

• One editor at a time. Whoever is editing holds the project; everyone else watches it change live, in their own window, as it happens. Ask for the pencil and it hands over in a click.
• Roles. Admins can invite people and change settings, editors can build, viewers can follow along. You choose who is allowed to publish.
• Suggest a change. Anyone watching can send the editor a prompt to run, so a colleague can ask for a fix without taking over the project.
• Share the result, not the project. Most people only need the published link. Publish once and send them the URL; they never need doable at all.

Invites are locked to the email you sent them to, so the person who opens the link has to sign in as that address. If they hit a dead end, the invite page offers a “Use a different account” way back.

What you can build

doable detects the right kind of project from what you ask for, and adapts the whole workspace to it, the preview, the publish flow, and the guidance the agent gets.

Databases & env vars

Most real tools need a database and some secrets. doable provisions these for you and keeps them in the Infrastructure tab, so the agent has a real backend to build against from the first turn.

• Databases. Attach a Postgres, Redis, or MongoDB to a project (or a single service) in one click. doable wires the connection string in as an env var automatically.
• Env vars. Set them per project or per service. Per-service values win over shared ones, so an API-only secret never leaks into the web app.
• Secrets you bring. The “Connect your services” panel collects third-party API keys (Stripe, OpenAI, etc.) and a banner flags anything still missing before you publish.

Doable Cloud & your own server

The app is free and building is unlimited, because it runs on the AI plan you already pay for. What costs money is publishing: every plan includes a monthly pot of credits covering builds, storage, and keeping your live apps running.

Brain

Brain is an optional memory layer: a private Postgres + pgvector store your agent can read from, so it already knows your project before you ask. It's also exposed over MCP, so other AI editors can use the same memory.

Read more about Brain →

Fixing problems

When a build or deploy fails, doable doesn't hand you a raw stack trace. It reads the logs, works out what went wrong (a missing env var, a lockfile drift, a broken build step), and explains the fix in plain English, often offering to apply it for you.

You can also run the doctor from the CLI in any project:

Still stuck? The support page walks common issues to a fix, and lets you message us if none of them match.

CLI & MCP

Everything in the app is also available from the terminal and from other AI editors. If you live in a terminal, or want your own editor's agent to publish without leaving the chat, install the CLI:

doable setup also writes the MCP config for Claude Code, Cursor, and Codex. After that, those editors pick up /deploy and the other doable tools as commands, so your AI can publish, read logs, and manage databases from inside the chat.

Databases reference

Databases attached in the app are also manageable from the CLI: back up, restore, and run schema migrations.

Account & billing

Your account is shared across the app, the dashboard, and the CLI. Manage your plan, usage, and payment details from the dashboard under Billing.

• The desktop app is free, with no limit on what you build locally.
• Plans include a monthly credit allowance for publishing, storage, and runtime. Credits are not AI usage, that stays on your own Claude, ChatGPT, or Gemini plan.
• Run out and published apps pause until the month resets, unless you turn on top-ups, which have a spending cap you set.
• Need help, or want to recover a deleted account? Head to Support.