How to integrate Gemini MCP with Grok Build

Connect Grok Build to Gemini MCP. Summarize this research article in 100 words, generate a creative image of a futuristic city, and more from your terminal, with authentication handled for you.

Gemini logoGemini
No AuthApi Key

Gemini is Google's multimodal AI platform for generating text, video, and rich chat completions. It's a one-stop shop for advanced AI content creation and intelligent automation.

8 Tools

How to integrate Gemini MCP with Grok Build

Grok Build is xAI's terminal coding agent. It runs on Grok 4.5, plans its work before it acts, and can run multiple sub-agents in parallel. It also reads Claude Code's MCP configuration, so any server you already have in a .mcp.json is picked up with no changes.

In this guide, I will show you how to connect your Gemini account to Grok Build through Composio, so it can summarize this research article in 100 words, generate a creative image of a futuristic city, create a 30-second video based on this script, and more without leaving the terminal. Composio holds the OAuth tokens for you, and Grok Build only calls the tools you approve.

For exact tool slugs and parameter tables, see the tool catalog in the Composio docs.

Also integrate Gemini with

Why use Composio over a standalone MCP server?

  • Read and write access. Composio's Gemini integration lets Grok Build take real actions like creating drafts, sending updates, and labeling records, not just reading data.
  • 1,700+ SaaS toolkits out of the box. One endpoint gives you a full catalog of pre-built connectors, from Gmail and Slack to Notion, Linear, and Salesforce.
  • One MCP server for every app. Wire up a single Composio server instead of maintaining a separate MCP entry for each app.
  • Smart, context-aware tool loading. Grok Build caps how many tools it holds in a single request. Composio loads only the tools a task needs, so you do not spend that budget on tools you are not using.
  • Cross-app automation. Chain actions across apps in one run. Pull a thread, summarize it in Notion, and post the highlights to Slack from a single prompt.

Prerequisites

  • Grok Build installed and signed in. Install with curl -fsSL https://x.ai/cli/install.sh | bash on macOS or Linux, or irm https://x.ai/cli/install.ps1 | iex on Windows PowerShell. On first launch Grok opens a browser to authenticate; for headless or CI use, set an XAI_API_KEY environment variable instead (create the key at console.x.ai).
  • Access to the Gemini account you want to connect.
  • The Composio MCP endpoint. Composio's server is remote and hosted, so there is nothing to run locally and no tunnel to set up.

Step-by-step: Connect Gemini to Grok Build

1. Install and verify Grok Build

Install the CLI, then restart your shell so the grok binary lands on your PATH:

bash
curl -fsSL https://x.ai/cli/install.sh | bash
which grok

On Windows, install with PowerShell instead: irm https://x.ai/cli/install.ps1 | iex. If which grok returns a path, you are set. Grok Build runs on Grok 4.5 by default; you can switch models inside the session with /model <name>.

2. Add the Composio server

Add Composio as a remote HTTP MCP server with the grok mcp add command:

bash
grok mcp add --transport http composio https://connect.composio.dev/mcp

You can also add and manage servers from inside a session. Run /mcps to open the extensions modal on the MCP tab, then add a new server and paste the Composio URL. Added this way, Grok auto-detects the name from the URL and lists the server as connect:

bash
https://connect.composio.dev/mcp
Grok Build /mcps extensions modal listing MCP servers Grok Build adding the Composio MCP server with its URL

Grok also reads Claude Code-style config, so an entry in ~/.grok/config.toml or a project .mcp.json works the same way.

3. Authenticate

Composio uses OAuth. In the /mcps modal, select the Composio server and press i to authenticate (Grok also triggers this browser flow automatically the first time it uses a Composio tool). Click Allow to authorize access. Grok stores the tokens under ~/.grok/mcp_credentials.json, and /mcps shows Composio as connected.

Grok Build prompting to authenticate the Composio MCP server Composio authorization screen with the Allow button for Grok Build

4. Start building

Ask Grok to work with your Gemini account through Composio. On the first Gemini action, Composio prompts you to connect the account through OAuth. Approve the scopes once, and Composio handles token refresh from there.

What you can do after connecting Gemini

  • Summarize this research article in 100 words
  • Generate a creative image of a futuristic city
  • Create a 30-second video based on this script
  • Find the most similar documents to this paragraph

Security + privacy notes (important)

  • Use least-privilege access. Grant only the Gemini scopes you actually need.
  • Review OAuth scopes before approving. Check that the requested scopes match what you expect Composio and Grok Build to do.
  • Keep write actions human-reviewed. Grok Build proposes a plan before it acts. Leave that approval step on for actions like sending messages or editing records.
  • Keep secrets out of version control. Your XAI_API_KEY and any tokens should never be committed. Use environment variables or a secrets manager.
TOOLS

Supported Tools

Every Gemini action and event your agent gets out of the box.

Count Tokens (Gemini)

Counts the number of tokens in text using Gemini tokenization.

Embed Content (Gemini)

Generates text embeddings using Gemini embedding models.

Generate Content (Gemini)

Generates text content or speech audio from prompts using Gemini models.

Generate Image (Nano Banana)

Generates a downloadable image from a text prompt, defaulting to Nano Banana 2 Lite on managed connections and Gemini 3 Pro Image Preview on other connections.

Generate Videos

Starts an MP4 video generation from a text prompt with Gemini Omni 1.

Get Video Generation Status

Checks a Gemini Omni Flash video generation started by GEMINI_GENERATE_VIDEOS once, without waiting.

List Models (Gemini API)

Lists available Gemini models with their capabilities and limits.

Wait and Download Video

Waits up to about 45 seconds for a Gemini Omni Flash video started by GEMINI_GENERATE_VIDEOS and returns the MP4 when ready.

FAQ

Frequently asked questions

Start with Gemini.It takes 30 seconds.

Managed auth, hosted MCP servers, and every Gemini tool your agent needs.Free to start.

Start building