> For the complete documentation index, see [llms.txt](https://docs.rockhopper.co/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.rockhopper.co/it-setup/mcp-server-getting-started.md).

# MCP Server - Getting Started

> Connect your AI assistant to your Rockhopper workspace so it can audit cell-level history, run review workflows, and version your spreadsheet work — without leaving your chat.

This guide gets you connected in about 2 minutes. It works with any AI assistant that supports the Model Context Protocol — including Claude (Desktop, claude.ai), ChatGPT, Cursor, Endex, and others.

There are two ways to connect:

* **Option A — Sign in with Rockhopper (recommended).** No setup; you sign in through your browser like any web app.
* **Option B — Use an access token.** A short copy-paste alternative if your AI client doesn't support sign-in yet.

{% hint style="info" %}
**Want to try it before connecting your AI client?** The [Rockhopper MCP Postman Workspace](/it-setup/mcp-postman-workspace.md) lets you click through every Rockhopper tool against your own account in your browser — no install required. Great for previewing what your AI will be able to do.
{% endhint %}

***

## Option A — Sign in with Rockhopper (recommended)

The simplest path: your AI client connects to Rockhopper's hosted MCP gateway and opens a browser for you to sign in — just like signing in to any web app. No software to install, no token to copy or store.

### Step 1 — Add the Rockhopper MCP server to your AI client

In your AI client's settings, look for "Add MCP server", "Connectors", "Custom Integrations", or "Remote MCP Server". Provide:

* **URL:** `https://mcp.rockhopper.co/mcp`
* **Transport:** Streamable HTTP
* **Authentication:** OAuth 2.0 (your client handles this automatically — no client ID or secret to enter)

Different clients label things differently, but every supported client just needs the URL above.

### Step 2 — Sign in

When you save the new MCP server, your client opens a browser window pointing to Rockhopper. Sign in with your normal account (the same one you use at [app.rockhopper.co](https://app.rockhopper.co)) and approve the connection.

That's it — your AI assistant is now connected.

### Step 3 — Test it

Ask: *"List my Rockhopper files."* You should see your enrolled workbooks come back.

ℹ **One caveat:** Option A doesn't work with a small number of desktop AI clients that use a local-loopback (`http://localhost:…`) OAuth callback during sign-in. If your client's first sign-in attempt fails with a 403 at the registration step, use **Option B** below instead.

***

## Option B — Use an Access Token

Use this if your AI client doesn't yet support the sign-in flow above, or if your IT environment needs the connection to run locally instead of through the hosted gateway. You'll create an access token in Rockhopper, then paste it into your AI client's settings.

### Step 1 — Create an access token

1. Sign in at [app.rockhopper.co](https://app.rockhopper.co).
2. Click your avatar (top right) → **Access Tokens**.
3. Click **Create token** and fill in:
   * **Name** — something memorable (e.g. *My AI Assistant*)
   * **Permission level** — *Read & write* to let your AI post comments and open reviews; *Read-only* if you only want it to read
   * **Expires** — 90 days is a good default
4. Click **Create**. Copy the token (it starts with `rh_pat_…`) — Rockhopper only shows it once. Keep it somewhere safe, like your password manager.

### Step 2 — Add Rockhopper to your AI client

Open your AI client's MCP settings (often labeled "MCP Servers", "Custom Integrations", or "Tools"). Add a new server with these values:

| Field                | Value                                                 |
| -------------------- | ----------------------------------------------------- |
| Name                 | *Rockhopper*                                          |
| Type                 | *Local* / *stdio* (whichever your client offers)      |
| Command              | `npx`                                                 |
| Arguments            | `-y @rockhopper-co/mcp-server`                        |
| Environment variable | `ROCKHOPPER_TOKEN` = *(paste your token from Step 1)* |

Save the settings. Your AI client may ask to install Node.js if it's not already present — that's normal and safe.

### Step 3 — Restart your AI client and test

Most clients only pick up new MCP servers at startup. Quit and relaunch, then ask: *"List my Rockhopper files."*

{% hint style="info" %}
**For developers configuring this via a JSON config file** (Claude Desktop's `claude_desktop_config.json`, Cursor's `mcp.json`, VS Code's `.vscode/mcp.json`, etc.):

```json
{
  "mcpServers": {
    "rockhopper": {
      "command": "npx",
      "args": ["-y", "@rockhopper-co/mcp-server"],
      "env": {
        "ROCKHOPPER_TOKEN": "rh_pat_paste_your_token_here"
      }
    }
  }
}
```

Requires Node.js 20 or newer on the machine running the AI client, and outbound access to `registry.npmjs.org`. See the [main MCP Server page](/it-setup/mcp-server.md) for more configuration details.
{% endhint %}

***

## What your AI assistant can do once connected

Example things to ask:

* **Audit cell-level history:** *"Who last changed C12 in the Q3 P\&L Model, and what was it before?"*
* **Review uncommitted changes:** *"Show me every unattributed change in our budget model this week and who made them."*
* **Comment and collaborate inline:** *"Add a comment on Sheet1!B5 asking why this assumption shifted 10%."*
* **Run formal approval workflows:** *"Create a review request for the latest version of the Q3 forecast."*
* **Approve, cancel, or chase reviews:** *"List my pending review requests, then approve the one for the budget model."*
* **Version, snapshot, or roll back:** *"Create a new version called 'pre-board-edit', then discard uncommitted changes on the Q3 forecast."*

The AI will only see what your Rockhopper account is allowed to see. If you give it a `read-only` token, it cannot post comments or open reviews even if asked.

***

## Security notes

* Treat your Personal Access Token like a password. Don't paste it into chat messages, commit it to Git, or email it.
* You can revoke a token at any time from the same **Access Tokens** page — the effect is instant.
* Rockhopper logs every action taken through MCP just like every other action in the app, so you'll see your AI's activity in your audit log.

***

## Help & support

* **Documentation:** [docs.rockhopper.co/it-setup/mcp-server](https://docs.rockhopper.co/it-setup/mcp-server)
* **Support:** <support@rockhopper.co>
* **Sandbox / try-before-installing:** there's a [Postman workspace](https://docs.rockhopper.co/it-setup/mcp-postman-workspace) you can click through to see every tool exercised against your account before you wire it into your AI client.

If you hit any issue — wrong tool behavior, OAuth errors, or anything that doesn't match this guide — email support and we'll get you unstuck. Please mention which AI client you're using; it helps us reproduce.