Skip to content

Introducing the SurfacedBy MCP Intelligence Server

Ali Khallad4 min readUpdated
June 20, 2026 , 4 min read
SurfacedBy product update hero: Bring your AI visibility into any agent, a read-only MCP server, beside the connection scopes panel
Share

If you run AI visibility for more than one brand, the same questions come back every morning. Which brand needs attention today? What changed this week? Which opportunity is ready to act on? What can I tell the client? Until now, answering them meant opening each dashboard, reading the cards, and copying the parts that mattered into whatever you were writing.

SurfacedBy now hands that data to your own AI tools. The MCP Intelligence Server is a read-only connection that lets an approved AI client read your brand’s visibility, cited sources, opportunities, and AI traffic directly. It works with Claude, ChatGPT, Cursor, and anything else that speaks the Model Context Protocol. You ask in plain language, the agent reads the live data, and you get the answer without leaving the tool you are already in.

Read-only by design

The server reads. It does not act. Every connection is scoped to exactly what you allow, carries its own key, and can be revoked the moment you want it gone. No matter which scopes you grant, a connection can never:

  • Post, publish, or send anything externally
  • Run a check or spend credits
  • Write the prompts your brand is checked against
  • Read secrets or another account’s data

That boundary is the point. An agent can summarize your week, but it cannot change your week. Business workspaces can optionally turn on a few reversible management actions, such as accepting a prompt suggestion or adding an opportunity to a plan, and even those stay inside the same rule: nothing publishes, nothing posts, nothing spends.

SurfacedBy Connect AI client panel showing read-only scopes for workspace, brand, visibility, opportunities, reports, and AI traffic, with a list of actions the connection can never do
Each connection is scoped to what you allow, and the things it can never do are spelled out before you connect.

What your agent can read

Scopes map to the surfaces you already use in the dashboard, so an agent reads the same trusted numbers rather than a separate copy:

  • Workspace overview. The portfolio view across your brands: the attention queue and per-brand summaries.
  • Brand basics. A brand’s overview, whether AI crawlers can reach the site, connected integrations, and how recent the latest check is.
  • Visibility analytics. Visibility Score, Brand Coverage, cited sources and competitors, tracked prompts, per-theme answer breakdowns, and how AI assistants describe the brand.
  • Opportunities. Recommended actions and source gaps, with the evidence behind each one.
  • Plan status, reports, and content. Where plan items stand, generated report summaries, and content projects.
  • AI-referral traffic. Visits and conversions from AI assistants, broken down by platform, with a trend.

Brands are addressed by domain name, the way you think about them. You ask about yourbrand.com, not an internal id, and internal ids never travel over the wire.

Set it up in about a minute

Open Settings, then Integrations, then AI clients, and choose Connect AI client. Name the connection, pick the scopes, and optionally restrict it to a single brand. SurfacedBy hands you a setup line with the key already in it. One key works across every client.

SurfacedBy MCP setup card with a Claude Code terminal command to add the read-only server at api.surfacedby.com, the endpoint URL, and example questions to ask an AI agent
A single read-only key connects Claude, ChatGPT, Cursor, or any MCP client.

For Claude.ai or ChatGPT, paste the server URL into the custom-connector field and leave authentication off. For Claude Code, Codex, Cursor, or VS Code, copy the one-line command or the mcp.json snippet. For anything else, the mcp-remote bridge covers any client that speaks the spec.

Questions worth asking

Once it is connected, the day-to-day questions answer themselves from live data:

  • What changed for yourbrand.com this week?
  • Which opportunities have the strongest proof behind them?
  • Summarize the source gaps for the client call.
  • Which brand in my portfolio is slipping against competitors?

For an agency, that last question is the useful one. Point an agent at the workspace scope and it can read the attention queue across every brand you manage, then draft the note you would have written by hand. The opportunities scope carries the same evidence the dashboard shows, so the agent can explain why an action matters, not just that it exists.

What it is, and what it is not

This is an intelligence layer, not an automation layer. The dashboard stays the source of truth: scoring, freshness, and roll-ups are computed on our side, and the MCP server reads the result rather than recomputing it. So an agent and the dashboard never disagree about a number. Connecting an AI client does not hand it the keys to your account; it hands it a window into your data, with the glass clearly marked.

The MCP Intelligence Server is available on Professional and Business plans under Settings, then Integrations, then AI clients.

A note on timing

SurfacedBy changes often. We build from a mix of customer requests, our own research and testing, and keeping pace as new AI clients and protocols appear, so the scopes, the client list, and the setup screens here reflect how the feature looked at launch. The tool catalog has grown since, and it will keep growing. If your client is not in the list above, the generic bridge almost certainly already covers it.