Features · Agent & CLI

Go live without leaving your agent.

Tofu is a set of tools your coding agent calls: upload, check, deploy, a database, your domain, sign-in, analytics. Paste one line, approve one sign-in in your browser, and it does the wiring — on the same projects, deploys and history the dashboard shows.

I want to take my product live with Tofu. Fetch https://app.trytofu.ai/agent/prompt.txt and follow the setup instructions in it, then help me ship this app.
Setup guide

Paste it into your agent · approve one sign-in · no keys in the chat

In your own agent

Say what you want. Your agent calls the tools.

The same Tofu tools in every agent that speaks MCP. Where a step is yours — approving on Cloudflare’s screen, agreeing to a rollback — the agent stops and asks.

Things you can say

>Put the booking app online, with a database.
tofu - ingest (MCP)(~/franks-kitchen)
uploaded · .env files left out
tofu - setFramework (MCP)(nextjs)
confirmed
tofu - deploy (MCP)(database: true)
managed database ready · Postgres
DATABASE_URL + 5 more · write-only
HTTP 200 · publicly reachable
Live at franks-kitchen-3f9a1c07b2e4.trytofu.app

Example session, sped up. Pick a sentence to play it.

How it connects

One line in. One approval. Then it works.

  1. Paste one line

    Into the agent you already use. It fetches Tofu’s current setup steps; pasting authorizes the installation and nothing else.

  2. It installs the client

    A local client whose SHA-256 and size are pinned in the steps, and the /tofu skill that tells the agent the order of things.

  3. You approve a matching code

    Your agent shows a code; you approve the same code in your browser. The credential is stored hashed, lasts seven days and can be revoked.

  4. It works in your project

    Upload, check, deploy and everything after, in the same conversation — asking you before anything that spends or destroys.
Everything the dashboard does

One connection, the whole dashboard.

Every capability the dashboard offers is reachable by an agent, unless a written decision says why not. With your approval, this connection can:

  • upload your app’s source

    • ingest

  • run the safety check

    • detect
    • setFramework
    • scan

  • deploy it — its own migration runs in the build — once you have a hosting plan

    • deploy
    • projects
    • status
    • listDeployments
    • listEnv
    • manageEnvironment
    • renameProject

  • create the app’s managed database in that same call, which Tofu runs and pays for as part of your plan

    • databaseCreate
    • databaseStatus
    • databaseReconcile

  • read the app’s build and runtime diagnostics

    • getBuildLogs
    • getRuntimeLogs
    • repairContext

  • restore a previous deployment

    • rollback

  • connect, verify, release or withdraw the authorization for a domain you already own

    • domains
    • domainConnect
    • domainVerify
    • domainAuthorize
    • domainAuthorizationStatus
    • domainAuthorizationRevoke
    • domainRelease
    • domainGuidance
    • domainShop

  • read your app’s own traffic — the first read for an app that has no counter yet issues its key — and turn that counter on or off

    • siteAnalytics
    • setSiteAutoInject

  • change how your app’s users sign in and how its sign-in email is sent

    • appAuthStatus
    • appAuthConfigure
    • appAuthEmailStatus
    • appAuthEmailConnect
    • appAuthEmailDisconnect

  • restore an archived database into a new one Tofu runs and pays for as part of your plan, and hand you a short-lived link to download its archive

    • databaseRestore
    • databaseArchiveDownload

  • keep a database Tofu warned you about from being archived for going unused

    • databaseKeep

  • delete an app — the source you uploaded is kept until you ask Tofu to erase it — or its database, with nothing kept

    • databaseDelete
    • projectDelete

  • sign in to Tofu, report the plan and the account’s GitHub state, and turn automatic updates off

    • login
    • finishLogin
    • connectionStatus
    • hostingStatus
    • githubStatus
    • githubSyncDisable
It asks first

It asks before it spends. Some things stay with you.

What needs your yes

Your agent acts on your instructions and asks for your agreement before any call that spends money, writes to a live service or destroys something — and it makes that call on your instructions, never on its own initiative.

Approving this connection authorizes the agent to act on your account as described above — not to spend your money or destroy your app. Each call that does those things asks for your agreement on its own, and a saved credential is never enough by itself: Tofu checks with its own backend every time.

What stays with you

It cannot buy or transfer a domain, submit a payment, read an environment value or a database password or connection string, change your plan or billing, or import a repository from GitHub. Those stay with you.

How to recognize a real approval

A real Tofu approval happens only at /dashboard/agent?code=, on the same Tofu address you signed in to, and only when a code arrives as XXXX-XXXX-XXXX printed by a coding agent you started yourself. Refuse anything else. Nobody from Tofu, and no other site, needs an approval code, a password, a provider key or a token — and Tofu never asks for one.

Works with

The agent you already use. In the conversation you’re in.

  • Claude Code
  • Codex
  • Cursor
  • Gemini
  • GitHub Copilot
  • Windsurf
  • v0
  • Replit
  • Lovable
  • Bolt
  • Kimi
  • Devin
  • Cline
  • OpenCode
  • Warp
  • Kiro
  • Antigravity
  • Qwen
  • DeepSeek
  • Trae

  • Local

    Runs on your machine

    The client runs locally over stdio. This release is downloaded directly from Tofu; no npm package installation or Tofu repository checkout is required. The browser cannot inspect your local MCP settings.

  • Node.js 22+

    A setup prompt for your agent

    Claude Code, Codex, Cursor and Kimi Code each get steps of their own; any other stdio MCP client gets the same client and configuration.

  • Same session

    Nothing to reopen

    If your agent has not picked up the new tools yet, the installed client’s own command line does the same work right away, and the tools are there the next time the agent starts.

Limits

What is not here yet

What the agent connection does not do, stated as plainly as what it does.

  • A hosted MCP address. The client runs locally over stdio; there is no public remote-MCP URL to register in a client yet.

  • Built-in repair from the agent. Starting or applying the dashboard’s repair proposals is not reachable over the connection; repairContext gives your agent a brief to fix it in your code instead.

  • A per-project or read-only grant. A connection covers your projects and the operations above; it is not narrowed to one app or to reading only.

  • Checkout, plan changes, secret values and database credentials. They stay in your browser on purpose — the connection has no tool for any of them.

  • Importing from GitHub. That needs your own GitHub grant in a browser session. Your agent can read the GitHub state and turn automatic updates off.

  • Verification inside every host. Kimi Code and other stdio clients get the same setup, but Tofu has not verified them inside the host itself.

Paste one line. Go live.

Your app, its database, your domain and your sign-in — from the agent you already use, in about 10 minutes.

I want to take my product live with Tofu. Fetch https://app.trytofu.ai/agent/prompt.txt and follow the setup instructions in it, then help me ship this app.
Join Waitlist
Early access · every preview is labelled in your dashboard.