Hosting for web apps.
Deploy and manage with the dashboard or Claude Code

It works the same with Codex, Cursor and any other agent.

Your agent creates the project, deploys, reads the logs and the errors your users hit, and fixes them. Nothing reaches production without your yes.

A session in Claude Code. Asked for a notes app, the agent builds it and runs it locally. Asked to put it online, it creates the project and gives a link to approve the deploy, then the address the app is live at. Asked about errors, it reports how many are open, fixes one, and asks for approval again. Asked how the app is doing, it reports requests and the slowest page. Asked for Google sign-in, it asks for approval and turns it on.

Agent-ready hosting

Connect your agent once. From then on it works in your account, with the permissions you gave it.

Works with the agents you already use

Any agent that speaks MCP connects the same way. There is one address to add and nothing to paste: your agent asks to connect, you sign in in the browser and choose what it may do.

See what your agent will read
  • Claude Code
  • ChatGPT Codex
  • Cursor
  • Gemini
  • Qwen Code
  • Any MCP client

MCP server. Recommended Your agent gets the whole platform as tools: projects, deploys, logs, errors, performance, sign-in, email and domains.

How connecting works

Command line. It needs no install, deploys the folder you are in, and answers in a form agents read as easily as people do.

See the commands

Docs as text. Every page of the documentation is served as plain text, so your agent reads the same contract you do.

Open the text version

What do you want your agent to do?

Ask in plain words. Bunderhost gives your agent the tools for it and keeps you in charge of what ships.

  • Deploy from a prompt

    It creates the project from your repository or a folder, prepares the database, storage and address, and deploys. You approve.

    Zero configurationOne-step rollback
    How deploys work
  • Fix the errors your users hit

    Errors arrive grouped, with the page and the release they broke in. Your agent reads the same list and the logs behind it.

    Grouped into issuesRuntime logs
    Running in production
  • Find what is slow

    The speed of every page and every request, measured on real traffic, so your agent starts from numbers and not from a guess.

    Real trafficEvery page
    See what is measured
  • Turn on sign-in

    Google and GitHub login without registering OAuth apps. One approval, and your agent switches it on and checks that it works.

    GoogleGitHub
    Sign-in and email
  • Send email from your domain

    Connect a domain, send mail and see what was delivered. Contacts and broadcasts are part of the platform.

    Delivery statusBroadcasts
    Sign-in and email
  • Preview every pull request

    Each pull request gets its own copy of the app, with its own database and file storage, before anything is merged.

    Own databaseSeed data
    Previews and domains

Every one of these works the same from the dashboard and from the command line. Choose where it runs

Nothing ships without your yes

An agent that can deploy must not be able to surprise you. Creating a project, deploying, reading your users’ data and changing sign-in each wait for a person.

What your agent asks for

Create the project notesAsked once, approved in the browser
Deploy to productionEach deploy, or a policy for one repository
Read the orders tableYour users’ data is a separate permission
Set a payment keyYou type it in the dashboard, not in the chat
See everything that needs you

Read only, or full access

  • You choose when the agent connects
  • Never more than your own role allows
  • Chosen for each organization
  • Revoked in one click

A deploy needs a yes

  • Confirm each one in the chat
  • Or approve a policy for one repository
  • An approval link waits 24 hours
  • Rollback follows the same rule

Your users’ data stays closed

  • Tables, files and jobs are opt-in
  • App users need their own approval
  • One environment at a time

Secrets never enter the chat

  • You type values in the dashboard
  • No tool hands a secret back
  • Separate tokens for CI, revoked at once
24 h
an approval link stays open, so you can answer later
0
secrets pass through the chat, in either direction

Connect the repo. That is the whole setup.

Built for JavaScript and TypeScript. Your code declares what the app needs, and Bunderhost reads it and prepares the database, file storage, jobs and address. There are no pipelines to write and no config files to keep.

  • It runs on your machine before you set anything up
  • Every push deploys, every pull request gets a preview
  • An earlier release is one step away
Read the docs

A dashboard for everything
your agent did

Deploys, logs, errors, performance, data and users in one place. You and your agent look at the same thing.

View all features

Deployments

Every release with its build, migrations and health, whoever shipped it. Any earlier one can go live again.

Logs

What the build printed and what the running app prints, searchable and kept apart.

Data and users

Browse the app’s tables and files, and see who signed up, without building an admin page.

Choose where it runs

On a server you already pay for, Bunderhost is free. Or let us run it and pay for what the app uses.

Your own server Free

Connect any Ubuntu or Debian server with a public address. You keep paying your provider for the machine, and Bunderhost adds nothing on top.

  • One command connects the server
  • The same dashboard, logs and agent tools
  • Databases and previews can live on it too
  • Its databases are backed up
  • HTTPS is set up for you
Connect your own server

Managed Pay for usage

Nothing to set up. Machines, database, file storage and domains are ready on the first deploy, and you pay for what the app actually uses.

  • No server to look after
  • Database and file storage included
  • Your own domains
  • Usage shown in the dashboard
  • Move to your own server later
Create an account

Onboard your agent with one prompt

Paste it into Claude Code, Codex, Cursor or any other agent. It reads the guide, connects to your account and takes the app from source to a live address.

See what your agent will read

Read https://your-bunderhost-domain/docs/agents and follow it to build and deploy my app.

Connect to my Bunderhost account by adding an MCP server named "bunderhost" at https://your-bunderhost-domain/mcp. It will ask you to authorize — open the link it gives you; I will sign in there and choose an organization.

Give your agent a host

One prompt to connect it. After that, shipping the app is a conversation.