API and MCP server
Your AI visibility data, in your own tools. On every plan.
A read-only REST API and an MCP server. Both run the exact code that draws your dashboard. Included with every plan, not locked behind an enterprise tier. Fourteen read endpoints, daily history, and a contract that only grows.
Plans with API access
All
the key's scope decides, not your plan. No enterprise gate
Read endpoints
14
13 on every tier, crawler analytics on Growth and up
MCP tools
16
every one reads, except run_fact_check
Contract
v1, grows only
released fields never change. X-MF-API-Version on every response
No enterprise gate
One key, no enterprise gate
Access depends on the key's scope, not your plan. Create a key under Settings, API keys, and read your workspace's data. Most competitors keep API access for enterprise contracts. Peec, for example, documents API keys as Enterprise-only (July 2026).
Keys come in two scopes. Full, for reads. And a narrow ingest-only scope for edge workers that can post crawler logs but never read workspace data. Every keyed request is metered at 120 requests a minute, with X-RateLimit-* headers and an honest 429 with Retry-After when you hit it.
-H "Authorization: Bearer mf_..."
"visibility": 61.4,
"share_of_voice_fraction": 0.312,
"citation_share": 0.208,
"sentiment_index": 64
}
Sample. The API runs the same code that draws the dashboard.
13 read endpoints
The same code that draws your dashboard
Brands, the visibility scorecard, prompts, competitors, cited sources and single cited URLs, raw answers, shopping placements and trend, the cached fact check, captured ads, metric history, and crawler analytics (Growth and up, same gate as the dashboard). Plus one ingest endpoint for streaming AI crawler logs. The API and the dashboard run the same functions, so they cannot drift apart.
GET /api/v1/metrics/history returns the daily series behind the charts. Blended across the search engines like the dashboard, or per engine, with competitor series on request. List resources paginate with cursors past the default caps. Requests without pagination get the exact frozen response.
Sample. Days with no collection are nulls with sample_size 0, never made-up zeros.
The contract
It only grows, it is versioned, and it is public
The v1 contract is frozen. Released fields never change meaning, scale or name. Every change adds. Each response carries an X-MF-API-Version header dated to the newest change, the same as the top entry of the public changelog. Metric math is versioned on its own, so a change in a computation can never quietly change a field you already parse.
MCP server
Ask your own data questions, from any AI agent
The MCP server turns your workspace into 16 tools over the Model Context Protocol (Streamable HTTP). Every one reads, except run_fact_check, which runs the metered fact check. Point Claude, Cursor or any MCP client at /api/mcp with your API key and ask: which prompts did we lose this week, which pages earned citations, what does GPTBot fetch on our site.
One of those tools is sentiment with receipts: how each engine talks about you and every rival you track, the counts behind the score, and the verbatim sentences the tone was read from. Ask “what do the engines hold against us, quote it” and you get the answer, not an adjective.
Tool descriptions carry the honesty rules. A null is missing data, never zero. One observed day is a snapshot, not a trend. So an agent gets the rules right where it reads the data.
Sample. 4 of the 16 tools. Each description carries the honesty rules an agent needs.
Setup
Connect an assistant from one screen
The MCP server page in the app prints your endpoint, mints a key, and fills that key into a paste-ready block for Claude Desktop, Claude Code, Cursor, or any MCP client. No hunting through docs for a host name, no editing a sample by hand.
A key reads exactly the projects you scoped it to, and nothing else. Rotate it and the old one keeps working for 24 hours, so a running agent never breaks mid-afternoon.
"mcpServers": {
"mentionflow": {
"type": "http",
"url": "https://app.mentionflow.ai/api/mcp",
"headers": { "Authorization": "Bearer mf_..." }
}
}The MCP server page prints this with your endpoint and your new key already in it.
Frequently asked questions
Does MentionFlow have an API?
Yes. A read-only v1 REST API over the same data your dashboard shows: brands, the visibility scorecard, prompts, competitors, cited sources and URLs, raw answers, shopping placements, the cached fact check, captured ads, metric history, and crawler analytics. Plus one write endpoint for sending AI crawler logs. The API and the dashboard run the same code, so a number in the API and the number on your screen can never disagree.
Is the API available on every plan?
Yes. Access depends on the API key's scope, not your plan. Any workspace can create a full-scope key under Settings, API keys, and read its data. There is no enterprise gate.
What is the MentionFlow MCP server?
An MCP (Model Context Protocol) endpoint that turns your data into 16 tools an AI agent can call, from list_brands and get_overview to get_metrics_history and get_sentiment, which comes with the quotes the tone was read from. Every tool reads, except run_fact_check, which runs the same metered fact check as the dashboard. Point Claude, Cursor or any MCP client at it with your API key and ask about your own data. Each tool's description carries the honesty rules, for example that a null is missing data, never a zero.
Can I pull historical data over the API?
Yes. GET /api/v1/metrics/history returns the daily series behind the charts: visibility, share of voice, citation share, or sentiment index. Windows of 7, 14, 28 or 56 days, or a custom range. Blended across the search engines like the dashboard, or one engine, with competitor series if you ask. Days with no collection are nulls with a sample size of zero, never made-up zeros.
Will API changes break my integration?
No. The v1 contract only grows. Released fields never change meaning, scale or name. Every response carries an X-MF-API-Version header with the date of the newest change, and every change is in the public changelog. Metric math is versioned on its own, so a change in a computation bumps a version instead of quietly changing a field.
Is there a Looker Studio connector?
Yes. A Looker Studio community connector, linked from the Reports hub. Build and share your own dashboards on top of your MentionFlow data.
The full reference is in the docs: API and MCP overview, resources, MCP server and the changelog. Pair the API with client reporting for the deliverables, or AI crawler analytics for the ingest side.
Build on your visibility data
Create a key in Settings, API keys, and make your first request in under a minute. On any plan.
Start 7-Day Trial