For agents

AI visibility your agent can act on.

Connect Recometrix to Claude Code, Cursor, or anything that speaks MCP. Your agent reads the exact answers where you are missing, picks the next executable fix, ships it, and the following scan checks whether the answer changed.

Run a free audit Read the API docs

1 line

to connect from Claude Code

8 tools

metrics, answers, sources, actions

Read-only

your agent can't break anything

Verbatim

every stored answer and cited URL

Connect

One key, one line.

Create an API key in Settings, paste one command, and your agent can pull everything the dashboard shows. Other MCP clients take the JSON config instead.

Claude Code

claude mcp add --transport http recometrix https://recometrix.com/api/mcp --header "Authorization: Bearer rmx_YOUR_KEY"

JSON for other MCP clients

{
  "mcpServers": {
    "recometrix": {
      "url": "https://recometrix.com/api/mcp",
      "headers": { "Authorization": "Bearer rmx_YOUR_KEY" }
    }
  }
}

Your agent asks what to fix

get_actions returns executable jobs, not advice. Each one names a single target, the output to produce, the steps, and the check that tells the agent when it is done.

It ships the fix with the evidence open

The verbatim answers, cited URLs, and source gaps behind each job are one tool call away. The agent edits your pages or drafts the outreach right in your repo, grounded in what the AI tools actually said.

The next scan grades the work

Every fix keeps the questions it targets. Later audits compare those questions with the rest of your tracking set, so the agent's work gets measured the same way yours would.

The toolbox

What your agent sees.

The same data behind every number in the dashboard, as MCP tools. Full input schemas are in the API docs.

Browse the schemas

list_projects

List the brands/projects tracked in this Recometrix workspace, with their ids.

get_visibility_metrics

Latest AI-visibility metrics for a project (visibility score, mention rate, top-3 rate, stability, competitor share) plus a 12-run history.

get_prompt_results

Prompt-level answers from the latest completed audit: which AI tools mentioned the brand, at what rank, with which cited domains. Includes the verbatim answers.

get_sources

Ranked citation-source opportunities from the latest completed audit, including citation gaps and source type.

get_actions

Prioritized action recommendations for a project (what to fix first), each backed by stored answers.

get_referrals

AI referral events and crawler rollups for a project.

get_deliverables

Generated deliverables for a project, including grounded markdown drafts.

get_methodology

Methodology metadata for the latest completed deep audit, including provider, model, sample, and scoring caveats.

Agent questions.

Which agents work with it?

Claude Code, Claude Desktop, Cursor, and any MCP client that can send an Authorization header. Everything the MCP server exposes is also on the REST API for scripts and CI.

What can my agent read?

Whatever the dashboard shows: visibility metrics, verbatim answers, citation sources, executable actions, AI referrals, generated drafts, and the methodology behind each audit. The server is read-only by design, so an agent can browse everything and break nothing.

Which plans include it?

API and MCP access is included on the Grow and Agency plans. Create a key in Settings; it is shown once and can be revoked any time.

Will this actually move my score?

An agent ships the same fixes a person would, just faster and without losing interest. The next audit re-measures the questions each fix targeted, so you see observed movement, never a promise.