Skip to main content

​
Gauge MCP

The Model Context Protocol (MCP) is an open standard that lets AI clients securely connect to outside tools and data. Gauge runs an MCP server, so any MCP-compatible client — Claude, Cursor, ChatGPT, Gemini, Windsurf, Zed, or your own tooling — can query your Gauge data and take actions on your behalf, grounded in your organization’s real numbers instead of guesses. Gauge MCP authenticates with OAuth 2.0. You point your client at the server URL, it discovers the authorization endpoint automatically, and you approve access from inside Gauge. There is no API key to create or token to paste.
MCP exposes the same data and actions you already have in the app, scoped to the organization you authorize. Connecting a client does not grant it anything you cannot do yourself.

​
What You Can Do With It

Once connected, your client can reach across Gauge’s tools. The main categories are:
  • AI visibility and citations. Read how often your brand and competitors are mentioned or cited across AI answers, track visibility and performance over time, and see which sources and pages drive citations.
  • Topics and prompts. Browse the prompt library Gauge tracks, read individual AI answers, and curate prompts — create, generate, move, and tag them.
  • SEO and keyword reporting. Pull organic rankings, search volume and difficulty, and the keywords any domain or page ranks for, including competitors.
  • GA4, Search Console, and PostHog reporting. Run GA4 and Google Search Console reports and query PostHog where those integrations are connected.
  • Ads reporting. Read Google Ads and OpenAI Ads account summaries, campaigns, ad groups, and keyword performance.
  • Content pipeline. Run the full content workflow — create a brief, research topics, generate an outline, write and review the article, manage images, and publish to a connected CMS as a draft, to staging, or live.
  • Memory. List, read, and edit the memory files Gauge keeps about your organization.
  • Action Center. List and open recommended actions, create and assign them, set due dates, and update their status.
  • Ask Gauge. Send a natural-language question to Ask Gauge and get an answer grounded in your data.

​
Connecting Gauge MCP

The MCP integration lives under Settings → Integrations → MCP. That page has setup details for the most common clients, plus generic connection details for anything else.
1

Open the MCP integration

Go to Settings → Integrations and open the MCP card. Pick the client you use, or scroll to the generic instructions if it is not listed.
2

Add Gauge to your client

For CLI clients, copy the ready-made command. Claude Code uses Copy install command to add the server, and Codex does the same. For Claude.ai, a Claude organization admin adds Gauge as a custom connector once using Copy connector URL, and members then click Connect next to Gauge.
3

Connect any other client

For anything else, point your client at the server URL https://app.withgauge.com/mcp over HTTP (Streamable HTTP or SSE). Use Copy server URL to grab it.
4

Authorize with OAuth

The first time you use the connection, your client redirects you to Gauge in the browser. Select the organization you want and approve access. No API key or manual token is required.
If you work across multiple client organizations, choose the right one during the OAuth step — the client’s access is scoped to what you authorize.

​
MCP vs API Keys

MCP is the recommended way to connect an AI client. It authorizes through OAuth, keeps you signed in without a token to store, and exposes Gauge’s tools in the structured form clients expect. API keys are meant for scripts and server-to-server automation that talk to the Gauge API directly. They require you to create and safeguard a secret and pass it on every request, which is a poor fit for interactive AI clients. Reach for an API key when you are writing your own code, and use MCP when you are connecting a client like Claude or Cursor. You can generate keys under Settings → Integrations → API Keys.