Connect your app

An agent app connects to Tensorant at one address, with a browser sign-in or, in some apps, a personal token. This page has the setup for Claude Code, Cursor, Codex, VS Code, Claude and ChatGPT. For a first connection in a few minutes, see the MCP quickstart.

The address agents connect to is:

https://mcp.tensorant.io/mcp

Settings → Agents shows the same address, with Copy address and the setup for each app below.

Before you start

• An account in an organization. Viewers can connect agents too, at Read access.
• An agent app that can use a remote server. ChatGPT also needs a plan or workspace that allows custom MCP servers, and on Claude Team and Enterprise plans an owner adds the connector first: see Claude and ChatGPT.
• To train or deploy, your organization needs a GPU provider and storage connected by an owner in Connections. An agent can see which connections are set up, never their values, and cannot change them.

Connect with browser sign-in

Browser sign-in is the simplest way to connect. You add Tensorant's address to your agent app, then approve the connection in your browser. No token to copy.

Claude Code

Run this in a terminal. Then type /mcp inside Claude Code and follow the steps to sign in in your browser.

claude mcp add --transport http tensorant https://mcp.tensorant.io/mcp

Cursor

Add this to ~/.cursor/mcp.json, or to .cursor/mcp.json in one project. Cursor signs you in through your browser.

{
  "mcpServers": {
    "tensorant": {
      "url": "https://mcp.tensorant.io/mcp"
    }
  }
}

Codex

Add this to ~/.codex/config.toml. Then run codex mcp login tensorant in a terminal to sign in in your browser.

[mcp_servers.tensorant]
url = "https://mcp.tensorant.io/mcp"

VS Code

Add this to .vscode/mcp.json in your project. VS Code signs in to remote servers through your browser the first time it connects.

{
  "servers": {
    "tensorant": {
      "type": "http",
      "url": "https://mcp.tensorant.io/mcp"
    }
  }
}

Claude

In Claude, open Customize, then Connectors, and choose Add custom connector. Paste https://mcp.tensorant.io/mcp into the MCP server URL field and add it, then sign in to Tensorant when Claude asks.

On Team and Enterprise plans, an owner adds it first under Organization settings, Connectors, and each member then clicks Connect.

ChatGPT

In ChatGPT, open Plugins, select the plus button and choose Add custom MCP server. Enter a name and a description, choose a public endpoint and paste https://mcp.tensorant.io/mcp, then set up authentication and sign in to Tensorant when ChatGPT asks.

Your plan or workspace settings must allow custom MCP servers.

Approve it in your browser

After you add the address, the app sends you to a Tensorant tab in your browser.

1. If you are signed out, sign in. The request is still there afterwards.
2. The page asks "Allow app to use Tensorant?". It names the app and says where it returns: a website, "an app on this computer", or an app by name, such as "cursor app". For an app Tensorant cannot check, it warns "Tensorant cannot verify who made this app. Allow it only if you started connecting it just now."
3. If you belong to several organizations, choose the Organization the agent works in. The page starts on the one you last worked in in the console on this browser; check that it is the right one.
4. Under Access, choose Read, Build or Full. See Access levels. Levels your role does not allow are unavailable and say "Your role allows less". The page starts on the lowest level the app asked for, or a lower one if your role allows less, and on Read when the app asked for nothing. Apps often ask for all three levels, so it usually starts on Read: choose more only if the agent needs it.
5. Choose Allow. The page says "You allowed app". Go back to the app: Return to the app takes you there, and you can close the tab. Choose Deny instead to refuse, and the app is told.

The agent now appears under Settings → Agents, with Browser sign-in under How. The app renews its own access while you use it, and each renewal moves the date under Expires to 30 days later. After 30 days without use the connection ends, and the app asks you to sign in again. You also get an email: see Email when an agent connects.

An agent works in one organization. To use another, connect it again and choose that organization.

Connect with a personal access token

Use a token when the app cannot open a browser, such as a script, a CI job or a remote machine, or when you want a fixed end date. A token works like a browser sign-in: it acts as you, in one organization, at one level, and it appears in the same list. Tokens are for Claude Code, Cursor, Codex and VS Code. Claude and ChatGPT connect with browser sign-in only.

Create the token first: see Create a token. It starts with ta_ and is shown only once, when you create it. The examples below show ta_... where your token goes.

Claude Code token setup

Run this in a terminal. Claude Code sends the token with every request.

claude mcp add --transport http tensorant https://mcp.tensorant.io/mcp --header "Authorization: Bearer ta_..."

Cursor token setup

Add this to ~/.cursor/mcp.json, or to .cursor/mcp.json in one project. The token sits in that file as plain text, so keep the file out of version control.

{
  "mcpServers": {
    "tensorant": {
      "url": "https://mcp.tensorant.io/mcp",
      "headers": {
        "Authorization": "Bearer ta_..."
      }
    }
  }
}

Codex token setup

Add the first part to ~/.codex/config.toml and run the second in your shell before you start Codex. Codex reads the token from that variable, so it stays out of the file.

# ~/.codex/config.toml
[mcp_servers.tensorant]
url = "https://mcp.tensorant.io/mcp"
bearer_token_env_var = "TENSORANT_TOKEN"

# In your shell
export TENSORANT_TOKEN="ta_..."

VS Code token setup

Add this to .vscode/mcp.json in your project. The token sits in that file as plain text, so keep the file out of version control.

{
  "servers": {
    "tensorant": {
      "type": "http",
      "url": "https://mcp.tensorant.io/mcp",
      "headers": {
        "Authorization": "Bearer ta_..."
      }
    }
  }
}

Next steps

• Tools: what an agent can do at each access level, how spending works, and how files get in.
• Access and tokens: access levels, browser sign-in or token, revoking and limits.
• Members: roles, which set how much access an agent can have.
• Connections: the GPU provider and storage an agent needs to train and deploy.