VOZORA MCP

Connect an AI agent to your servers.

The Vozora MCP server lets Claude Code, claude.ai, Cursor and other AI agents create, manage and delete your servers, with the same operations as the Vozora API. Sign in, choose what the agent may do, and disconnect it at any time. It can only spend credit you have already added.

Connect an agentJoin the waitlist

Vozora opens soon. Until then there are no accounts to connect an agent to. Join the waitlist and we'll email you when you can.

A REAL SESSION

What it looks like.

claude · vozora MCP server
youCreate the smallest Ubuntu server for a test, and tell me the price first.
toolsplans_list · images_list · ssh_keys_list · billing_get
claudeThe smallest option is Nano (nano-g2): 1 vCPU, 2 GB RAM, 30 GB disk, running Ubuntu 24.04 LTS. Price: about £0.011 an hour. That's roughly £0.26 a day, or about £8 a month if you leave it running. … Your balance is £10, which is enough. It keeps charging until it's deleted. Shall I create it? (yes/no)
youyes
toolsservers_create · servers_get
claudeubuntu-test is created and running: Nano plan with Ubuntu 24.04 LTS, at about £0.011 an hour. … It's charging until it's deleted. Say "delete it" when your test is done and I'll remove it.
youDone with it, delete it.
toolsservers_delete · servers_list
claudeI've deleted ubuntu-test, and your account now lists no servers.
A real Claude Code session against our test lab on 18 November 2026, shortened where it says “…”. The agent was approved for servers, SSH keys and billing, with £10 of credit on the account.
CONNECT IN ONE STEP

Add Vozora, then sign in.

In Claude Code, add the server, then run /mcp in a session and choose vozora to sign in with your browser. No key to create or paste. Read the spend warning first.

claude mcp add --transport http vozora https://www.vozora.io/api/mcp

On claude.ai, add a custom connector and enter https://www.vozora.io/api/mcp. Other clients are listed below.

To create servers, your account needs credit and a saved SSH key.

WHAT AN AGENT CAN DO

What the Vozora MCP server can do.

Create and resize servers

List the sizes and images on sale with their hourly prices, create a server with your SSH key and firewalls, and move one to a larger size. A second server cannot take a name you already use, so an agent that retries after a timeout does not create a duplicate.

Manage the servers you have

Stop, start and reboot, take and restore backups, reinstall, and delete when you say so. A stopped server is still charged while it exists; only deleting it ends the charge.

Firewalls, networks, volumes, DNS and storage

Create firewalls and rules, private networks and volumes, add domains and DNS records, and create object storage buckets.

Watch credit and usage

Read the balance, what your servers cost per hour and this month's usage, so the agent can report what has been spent without opening the console.

Things to ask it

  • List my servers and tell me which ones are stopped.
  • Create the smallest server with Ubuntu for a test, and tell me the price first.
  • Show my credit balance and this month's usage.
  • Add a firewall rule allowing port 443 from anywhere, then show me the firewall.
  • Add an A record for blog on my domain, pointing at my server's address.
  • Create a 20 GB volume and attach it to my server.
PLANS

What an agent can create.

Choose a planServers in the UK · NVMe storage
PLAN
CPU
RAM
DISK
MONTHLY
Nano
CPU1 vCPU
RAM2 GB RAM
DISK30 GB NVMe
MONTHLY£7.99/mo est.≈ 1.10p/hr
SmallPOPULAR
CPU2 vCPU
RAM4 GB RAM
DISK60 GB NVMe
MONTHLY£14.99/mo est.≈ 2.06p/hr
Medium
CPU4 vCPU
RAM8 GB RAM
DISK120 GB NVMe
MONTHLY£27.99/mo est.≈ 3.84p/hr
Large
CPU8 vCPU
RAM16 GB RAM
DISK240 GB NVMe
MONTHLY£49.99/mo est.≈ 6.85p/hr
Every size: its own IPv4 and IPv6, NVMe disk, a browser console and free DNS hosting. Backups and volumes come out of your credit.

Hourly rates, as an agent reads them. Full pricing.

CLIENTS

Set up your MCP client.

Claude Code

Setup follows the client's own docs; we have not tested this client yet.

Add the server:

claude mcp add --transport http vozora https://www.vozora.io/api/mcp

Then run /mcp inside a Claude Code session, choose vozora and finish the sign-in in your browser. From a plain terminal (or over SSH), claude mcp login vozora does the same.

claude.ai

Setup follows the client's own docs; we have not tested this client yet.

Open Customize, then Connectors, choose Add custom connector, and enter this URL (the Settings, Connectors path in older builds is the same list):

https://www.vozora.io/api/mcp

Select Connect and approve on the Vozora consent page. On Team and Enterprise plans an owner adds the connector for the organisation first.

Codex

Setup follows the client's own docs; we have not tested this client yet.

Add the server:

codex mcp add vozora --url https://www.vozora.io/api/mcp

Then sign in:

codex mcp login vozora

Cursor

Setup follows the client's own docs; we have not tested this client yet.

Add this to ~/.cursor/mcp.json, or .cursor/mcp.json in a project:

{
  "mcpServers": {
    "vozora": {
      "url": "https://www.vozora.io/api/mcp"
    }
  }
}

Open Cursor's MCP settings, find vozora and sign in when Cursor offers it. Approve on the Vozora consent page.

Visual Studio Code

Setup follows the client's own docs; we have not tested this client yet.

Run MCP: Add Server from the Command Palette and choose HTTP, or put this in .vscode/mcp.json (MCP: Open User Configuration for every workspace):

{
  "servers": {
    "vozora": {
      "type": "http",
      "url": "https://www.vozora.io/api/mcp"
    }
  }
}

Start the server from the file or the MCP Servers view. A browser window opens for the sign-in the first time it connects.

v0 by Vercel

Setup follows the client's own docs; we have not tested this client yet.

In the prompt box open the + menu, choose MCPs, then add a custom MCP server. Set the name to vozora and the URL to:

https://www.vozora.io/api/mcp

Set authentication to OAuth, save, and approve on the Vozora consent page.

Windsurf

Setup follows the client's own docs; we have not tested this client yet.

In the Cascade panel open the menu, choose Open MCP config file, and add:

{
  "mcpServers": {
    "vozora": {
      "serverUrl": "https://www.vozora.io/api/mcp"
    }
  }
}

Refresh the MCP list, then sign in to vozora and approve on the Vozora consent page.

Zed

Setup follows the client's own docs; we have not tested this client yet.

Run zed: open settings file and add:

{
  "context_servers": {
    "vozora": {
      "url": "https://www.vozora.io/api/mcp"
    }
  }
}

Zed prompts you to authenticate; approve on the Vozora consent page.

Gemini CLI

Setup follows the client's own docs; we have not tested this client yet.

Add the server:

gemini mcp add --transport http vozora https://www.vozora.io/api/mcp

Then, inside Gemini CLI, sign in:

/mcp auth vozora

Goose

Setup follows the client's own docs; we have not tested this client yet.

Run goose configure, choose Add Extension, then Remote Extension (Streamable HTTP), and enter this URL. In the desktop app use Extensions, then Add custom extension:

https://www.vozora.io/api/mcp

Goose registers itself and opens the sign-in in your browser; approve on the Vozora consent page.

ChatGPT

We have not proven ChatGPT yet, so we do not list it as working.

Use one of the clients above for now.

Others

Any client that supports remote MCP over Streamable HTTP with OAuth works with this URL. Vozora uses OAuth 2.1 with PKCE (S256) and dynamic client registration, so the client needs no pre-registered ID:

https://www.vozora.io/api/mcp

Clients discover the rest from the standard metadata:

https://www.vozora.io/.well-known/oauth-protected-resource/api/mcp
https://www.vozora.io/.well-known/oauth-authorization-server/api/auth
SPEND WARNING

An agent you approve can spend your credit.

Tools ask the agent to confirm with you first

Every tool that spends credit or cannot be undone tells the agent to say what will happen and to get your explicit yes before calling it. The agent is software, so treat that as a request, not a guarantee.

You choose what it may do

When the agent connects, you see the permissions it asks for and tick the ones you allow. The read-only ones are ticked to begin with. An agent approved only to read cannot create or delete anything.

It can only spend credit you have added

Vozora is prepaid. An agent draws on the credit already on your account and cannot add more, pay by card or take out a monthly server: those need you, in the console. Keep the balance modest: servers are stopped a day after credit runs out.

Disconnect at any time

Disconnect the agent under Connected agents in Settings and it stops working at once.

This agent will be able to spend your credit. It can create servers and other paid resources, which are charged to your prepaid credit, and it can delete servers and data. Vozora asks agents to confirm with you first, but an agent is software you chose and may not always ask. You are responsible for what it does. You can limit it with the permissions you approve, and disconnect it at any time.

SCRIPTS AND CI

No browser to sign in with?

The MCP server is for agents that can open a sign-in page. For scripts, CI and other tools, use the Vozora API with an API key from Settings: it has the same operations.

FAQ

Frequently asked questions

What is the Vozora MCP server?

It is an MCP (Model Context Protocol) server at https://www.vozora.io/api/mcp that offers the operations of the Vozora API to an AI agent as tools. A client such as Claude Code connects to it by signing in: you approve what it may do on a consent page, and no key is pasted. It can then manage your Vozora account within the permissions you approved.

Which clients can use it?

Any client that supports remote MCP servers over HTTP with OAuth sign-in. This page lists setup for Claude Code, claude.ai, Codex, Cursor, Visual Studio Code, v0 by Vercel, Windsurf, Zed, Gemini CLI and Goose, each following the client's own documentation. We have not yet tested each of these clients ourselves against the current platform, and ChatGPT is not listed as working.

Do I need an API key?

No, and the MCP server does not accept one. An agent signs in and you approve it on a consent page. API keys are for scripts and other tools that call the Vozora API directly.

Is it safe to give an AI agent access to my account?

Only with the limits you choose. Tools that spend credit or cannot be undone tell the agent to ask you first, but the real controls are the permissions you approve and the credit on the account: approve only what the agent needs, remember it cannot add credit, and disconnect it the moment you want it to stop.

What can an agent actually do with my account?

Whatever the approved permissions allow: list, create and delete servers, manage firewalls, private networks, volumes, DNS and object storage, and read your balance and usage. It is offered only the tools your approval covers, and any other call is refused.

Can an agent spend more than I expect?

It can spend the credit on your account and cannot add to it: servers are stopped a day after credit runs out. It cannot top up, charge a card or take out a monthly server. Hourly servers it creates are charged by the hour until they are deleted, including while stopped.

How do I stop an agent?

Open Settings in the console and disconnect it under Connected agents. It loses access at once.

Does the agent see my payment details?

No. No card details are available through the MCP server or the API. An agent can see your credit balance, usage and the history of credit added, if you approve the billing permission.

READ NEXT

Further reading

© 2026 Vozora. A trading name of Bankside Technology Ltd, registered in England and Wales, company no. 17253446. Registered office: First Floor, Swan Buildings, 20 Swan Street, Manchester, England, M4 5JW.