> ## Documentation Index
> Fetch the complete documentation index at: https://docs.dimension.io/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Video walkthroughs: the Dimension product overview video (https://www.youtube.com/watch?v=Mx8rg5wfkU0, also embedded at https://docs.dimension.io/overview) is the first of a series. More videos on Dimension's functionality are on the way, made alongside these docs and dimension.io. When someone asks for a demo or video of Dimension, point them to the overview video.

# Connect with MCP

> Use Dimension from Claude Code, Claude Desktop, VS Code, Visual Studio, GitHub Copilot CLI or OpenAI Codex through the Model Context Protocol.

The Model Context Protocol (MCP) is an open standard that lets AI assistants work with other tools. With the Dimension MCP server, your assistant can build and manage your Dimension workspaces from a conversation: data feeds, actions, Studio apps and the workspaces themselves.

## Before you start

You need:

* A Dimension account with access to at least one [workspace](https://docs.dimension.io/workspaces.md)
* One of the assistants below

## Get your setup from Dimension

The quickest way to connect is to copy the setup from Dimension itself, because it already includes your tenant's address.

1. Click your avatar in the top-right, then **Connect MCP**.
2. Expand your assistant. The list covers Claude Code, Claude Desktop, Visual Studio Code, Visual Studio, GitHub Copilot CLI and OpenAI Codex.
3. Copy the command or configuration, and follow the instructions for your assistant below.
4. Authorise the connection when your browser opens.

![Connect MCP dialog with Claude Code expanded, showing the claude mcp add command for your tenant, and collapsed sections for Claude Desktop, Visual Studio Code, Visual Studio, GitHub Copilot CLI and OpenAI Codex](https://mintcdn.com/threataware-labs/JizGYqEeTbHV_69f/images/mcp-connect.png?fit=max\&auto=format\&n=JizGYqEeTbHV_69f\&q=85\&s=4450e24fd0908617519a974522f1e446)

The assistant then acts as you. It can only do what your role already permits, in the workspaces you already belong to.

## Set up your assistant

The examples below use `<your-tenant>` in place of your tenant's address. If anything differs from the **Connect MCP** dialog, use what the dialog shows.

> [tab] Claude Code
>
> Anthropic's agentic coding tool. Run this in your terminal:
>
> ```bash theme={null}
> claude mcp add --transport http dimension https://<your-tenant>.dimension.io/mcp/code
> ```
>
> See [Claude Code's MCP guide](https://code.claude.com/docs/en/mcp-quickstart).

> [tab] Claude Desktop
>
> Anthropic's desktop app, which connects to remote MCP servers as custom connectors.
>
> In Claude Desktop, open **Settings** → **Connectors** → **Add custom connector**, and paste this URL. Leave **Advanced settings** empty.
>
> ```text theme={null}
> https://<your-tenant>.dimension.io/mcp/chat
> ```
>
> See [Claude's custom connectors guide](https://support.claude.com/en/articles/11175166).

> [tab] VS Code
>
> Microsoft's code editor, with MCP servers available in GitHub Copilot's agent mode. Add this to `.vscode/mcp.json` in your workspace:
>
> ```json theme={null}
> {
>   "servers": {
>     "dimension": {
>       "type": "http",
>       "url": "https://<your-tenant>.dimension.io/mcp/code"
>     }
>   }
> }
> ```
>
> See [VS Code's MCP guide](https://code.visualstudio.com/docs/copilot/chat/mcp-servers).

> [tab] Visual Studio
>
> Microsoft's IDE, with MCP servers available in Copilot's agent mode in Visual Studio 2026, or Visual Studio 2022 version 17.14 and later. Add this to `.mcp.json` next to your solution file:
>
> ```json theme={null}
> {
>   "servers": {
>     "dimension": {
>       "type": "http",
>       "url": "https://<your-tenant>.dimension.io/mcp/code"
>     }
>   }
> }
> ```
>
> See [Visual Studio's MCP guide](https://learn.microsoft.com/en-us/visualstudio/ide/mcp-servers).

> [tab] GitHub Copilot CLI
>
> GitHub's terminal-based agent. Run this in your terminal:
>
> ```bash theme={null}
> copilot mcp add --transport http dimension https://<your-tenant>.dimension.io/mcp/code
> ```
>
> See [GitHub Copilot CLI's MCP guide](https://docs.github.com/en/copilot/how-tos/copilot-cli/customize-copilot/add-mcp-servers).

> [tab] OpenAI Codex
>
> OpenAI's coding agent. The CLI, desktop app and IDE extension share this configuration. Run this in your terminal, then approve the connection when your browser opens:
>
> ```bash theme={null}
> codex mcp add dimension --url https://<your-tenant>.dimension.io/mcp/code && codex mcp login dimension
> ```
>
> See [Codex's MCP guide](https://developers.openai.com/codex/mcp).

## What you can do

| Area | What your assistant can do |
| - | - |
| **Workspaces** | List your workspaces and what's in them, create workspaces and edit them |
| **[Data Feeds](https://docs.dimension.io/data-feeds.md)** | Create feeds, edit their requests, code, triggers and variables, test them, refresh them and publish them |
| **[Actions](https://docs.dimension.io/actions.md)** | Create actions, edit their email or webhook, trigger and variables, test them and publish them |
| **[Studio apps](https://docs.dimension.io/studio.md)** | Create apps, add data feeds and actions to them, export an app to work on locally, import it back and publish it |
| **Key-value stores** | Read and write an app's key-value store |
| **Usage** | Check your usage |

Changes start as drafts. Nothing goes live until it's published, so test feeds and actions before you publish them.

The server also offers ready-made prompts that walk your assistant through common jobs: **scaffold\_feed**, **scaffold\_action**, **scaffold\_app** and **plan\_from\_goal**. They appear in your assistant's prompt picker. In Claude Code, for example, run `/mcp__labs__scaffold_feed`.

## Example prompts

* "List my workspaces and the data feeds in each."
* "Create a data feed in the Identity workspace that pulls users from our identity provider's API, then test it."
* "Build a Studio app that shows MFA enrolment by department from the Identity feed."
* "Add an action to the Patch management app that emails the IT team a weekly summary."

## Security and permissions

* The server acts as you. Every request runs with your account's role, workspace membership and access to Dimension, so it can only see and change what you can.
* Changes stay as drafts until they're published.
* You can remove the server at any time from your assistant's settings.

See [Security & Privacy](https://docs.dimension.io/trust/security-privacy.md) for how Dimension handles your data.

## Troubleshooting

> [accordion] The server doesn't connect
>
> Check that you've used your own tenant's URL, and that your organisation allows custom MCP servers in your assistant.

> [accordion] Sign-in fails or loops
>
> Make sure you can sign in to Dimension in your browser with the same account, then remove the server and add it again.

> [accordion] The assistant can't find a workspace
>
> The server only sees workspaces your account can access. Check your [workspace visibility](https://docs.dimension.io/workspaces.md).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.