Connect Lumos to Claude via MCP: Native MCP AI Visibility

What is the Model Context Protocol (MCP)?

The Model Context Protocol is an open standard, released by Anthropic in late 2024, for connecting AI assistants to external data sources. Instead of copy-pasting reports into a chat window, MCP lets an AI agent call live tools — read data, query an API, fetch documents — directly inside the conversation. Anthropic is the host: Claude was the first assistant to ship native MCP support, and it remains the smoothest MCP experience as of 2026.

For a product like Lumos, MCP is the cleanest way to make AI visibility data usable: instead of asking your marketing team to log in and export a CSV, you ask Claude directly and it pulls fresh data from Lumos in the same conversation.

Why connect Lumos to Claude?

When the Lumos MCP server is connected, Claude can read your live workspace data inside any conversation. Claude is the natural home for MCP because Anthropic built the protocol — the setup is fastest, the surface is most polished, and the agentic patterns are most reliable. A few workflows light up immediately:

• Morning briefings. Ask Claude "summarize my Lumos visibility this week" and it pulls daily scores, top movers, and competitive deltas without leaving the chat.
• Ad-hoc competitive checks. "How am I doing against Stripe in Perplexity for SaaS payment prompts?" — Claude queries Lumos's competitor data and answers in plain English.
• Content workflows. When you're drafting a page with Claude, you can ask it to incorporate Lumos's recommended actions for that page — improving citability while you write.
• Agentic loops. Claude's agentic features (subagents, tool chaining, file operations) compose naturally with the Lumos MCP server — useful for multi-step audits where Claude reads Lumos data, drafts a fix, and writes it to a file.

How the connection works

The setup takes about two minutes — faster than ChatGPT or Gemini because Claude was built for MCP from day one.

1. Sign up for Lumos Pro. Start with the signup flow. The free tier doesn't include the MCP server — Pro and above do, and Pro starts with a 14-day no-credit-card trial.
2. Generate an MCP API key. In your Lumos dashboard, open Settings → Integrations → MCP. Click "Generate key" and copy the value. The key is scoped to your current workspace and is read-only.
3. Add the server to Claude. In Claude Desktop, open Settings → Connectors → "Add server", paste the Lumos MCP URL (provided in your dashboard alongside the key) and the API key. In Claude.ai, the same flow is in Settings → Tools → Custom MCP. In the Anthropic API, register the server via the MCP client library before your first message.
4. Test the connection. Start a new Claude conversation and ask: "summarize my Lumos visibility this week". Claude will call the MCP server, fetch your latest scan, and answer with live data.

What data is exposed via MCP?

The Lumos MCP server exposes the read-only surface of your dashboard, mapped to MCP tools that Claude can call:

• Daily visibility score per engine (ChatGPT, Gemini, Claude, Perplexity, Google AI).
• Mentions over time, including sentiment and the themes AI engines associate with your brand.
• Competitor head-to-head — who outranks you on which prompts, and by how much.
• The recommended actions backlog — Lumos's prioritized weekly list of pages to fix.
• Page-level citability scores for every page Lumos has graded.

What is not exposed: account settings, billing, team management, or any write action. The MCP server cannot change your Lumos configuration — it can only read.

Security notes

MCP API keys are read-only by default. Each key is scoped to a single Lumos workspace, so if you have separate workspaces for staging and production, you can issue separate keys for each. Keys are revocable from the dashboard with one click — revoking immediately disconnects any Claude (or other MCP client) using that key. We also log every MCP call to the audit trail in your workspace, so you can see exactly which queries were made and when.

Other MCP-friendly LLMs

The Lumos MCP server works with any MCP-compatible client. Today that includes Claude (this page), ChatGPT (see Connect Lumos to ChatGPT), Gemini, and a growing number of agentic frameworks. The same API key works across all clients — you don't need a separate key per LLM. Claude remains the most polished implementation because Anthropic owns the protocol.