MCP server

What the MCP server is

Analyze AI ships a remote MCP server that plugs your AI-visibility data into Claude, ChatGPT, and any MCP-compatible assistant — with evidence-backed, chart-ready answers.

Your data, inside your assistant#

Analyze AI ships a remote MCP server. MCP — the Model Context Protocol — is the open standard AI assistants use to talk to external tools. Once your workspace is connected, your assistant can query your real visibility data and answer questions like:

"I want visibility over the past 14 days, and sentiments, and a perception map."

What it lets you do#

• Run any analysis, on demand — more than 25 ready-made analyses covering share of voice, visibility movers, competitor gaps, sentiment alerts, citation quality, source authority, and daily trend series.
• Get finished deliverables — built-in playbooks produce complete outputs in one request: a weekly visibility brief, a competitor radar, a sentiment deep dive, a citation outreach plan, or a before/after campaign analysis. See example prompts.
• See charts, not walls of numbers — analyses declare how they should be visualized, so assistants that render charts turn your data into line, bar, pie, and scatter visuals automatically.
• Manage prompts conversationally (optional write access) — add or retire tracked prompts and trigger fresh runs from the chat, with plan limits enforced automatically.

What makes it different#

Most analytics integrations expose raw API endpoints and hope the assistant stitches them together. The Analyze AI MCP server is designed the other way around:

• Every claim carries receipts. Analyses return evidence references, and a dedicated tool dereferences them into the exact quotes, sentiment spans, and source snippets behind every number — with the engine name and run date. If a claim has no evidence, the assistant is told to say so rather than present it as fact.
• One question, one finished analysis. Instead of low-level endpoints, the server exposes curated analyses that each answer a real marketing question in a single call — "which high-authority pages cite competitors but not us?" is one tool call, not an afternoon of query planning.
• Playbooks built in. The weekly brief, competitor radar, and citation gap plan ship as ready-made workflows your assistant can run by name — your first session can produce a board-ready deliverable.
• Chart-ready by design. Time series carry their sample sizes, breakdowns carry their percentages, and every chartable result says what kind of chart it wants.
• Writes with guardrails. Prompt management is opt-in, owner-controlled, quota-checked before anything is written, and rate-limited — an over-eager agent can't burn your plan.
• Workspace-true access. Connections are scoped to one workspace, honor your team roles on every request, and can be revoked instantly.

Where to go next#

• Set up Claude & ChatGPT — connected in a couple of minutes.
• Example prompts — copy-paste questions that get great answers.
• Supported assistants & agents — what works with it, and how each connects.
• Limits & what it can't do — the honest list.