Connect Your AI via MCP (ChatGPT, Claude & More)
MCP (Model Context Protocol) is the open standard that lets AI assistants connect to external tools. LLM Pulse ships an MCP server, so you can ask ChatGPT, Claude, Gemini, Cursor, Windsurf, Zed, or any MCP-compatible client questions about your own visibility data, in your own AI tool.
With the LLM Pulse MCP connection, you can:
- Ask ChatGPT or Claude about your visibility, citations, and competitors without leaving your AI tool
- Connect once with OAuth, free on every plan, with no API key to manage
- Let your AI assistant pull live LLM Pulse data into any analysis or report it writes for you
- Use the same data from Cursor, Windsurf, Zed, Gemini, and any other MCP-compatible client
The OAuth-based connection is free on every plan and during eligible trials. Setup lives at User Menu > Connect AI (MCP) in the app.
What it does
- Exposes your LLM Pulse projects to your AI assistant through roughly 50 tools: visibility metrics, timeseries, share of voice, mentions, citations, sources, sentiment, prompts, responses, recommendations, and more
- Your assistant can also act: create prompts, add competitors, create tags, launch recommendation runs (it will ask you before creating anything)
- Access is scoped to your account and respects your plan: tools for features your plan does not include are simply not offered to the client
How to set it up
- In LLM Pulse, open User Menu > Connect AI (MCP) for the guided instructions per client
- In your AI client's connector or MCP settings, add the server URL:
https://api.llmpulse.ai/api/v1/mcp - Your client opens a browser window; sign in with your LLM Pulse account and authorize
- Ask something like "list my LLM Pulse projects" to confirm the connection works
For headless or scripted use (CI, automations), MCP also accepts a Bearer-token API key instead of OAuth. API keys require a Scale plan or above and are created under User Menu > API Keys.
Troubleshooting
- Connection fails or tools don't appear: remove the LLM Pulse connection from your AI client, re-add the URL, and complete the browser authorization again. A previously failed or expired authorization is the most common cause.
- Prefer OAuth over an API key for personal use; it is simpler and works on every plan
- A tool seems missing: tool availability follows your plan and, for team members, your permission role. If a feature is locked in the app, its MCP tool is filtered out too.
- Still stuck? Open a ticket from the support page with the client you are using (ChatGPT, Claude Desktop, claude.ai, Cursor) and what error you see
Tips & notes
- Your OAuth connection stays authorized for one year. After that, your AI client will ask you to authorize it again.
- MCP works well for ad-hoc questions and cross-tool workflows; for in-app analysis with charts, the built-in LLM Pulse Agent is faster
- The REST API (Scale and above) remains the right choice for building dashboards and custom integrations; see the REST API overview