Mentionfront runs a standard MCP server, so Claude, Gemini CLI, Cursor, ChatGPT or any MCP client that supports a bearer-token connection can search your mentions, pull the Opportunities and Needs attention queues, and tag or resolve mentions, all without opening the dashboard. This page covers what it can do, the exact tools it exposes, and how to connect one.
No credit card. MCP server included on every plan.
MCP (Model Context Protocol) is the open standard AI agents use to connect to outside tools and data. A social listening MCP server puts your mentions inbox behind that same standard, so an agent like Claude or Cursor can query and act on it directly, the same way it reads a file or calls an API you've configured for it.
For a listening tool, that turns two kinds of work into a chat message instead of a dashboard visit:
Mentionfront's MCP server is a real implementation of this, not a wrapper around a generic API: sales opportunity detection, the Needs attention queue and the priority score behind them are the same ones the tools below read from, so an agent's answer matches what you'd see in the inbox.
Every tool is scoped to your workspace and described with its own input schema, so a client reads what's available and how to call it the moment it connects — nothing to configure by hand beyond the connection itself.
| Tool | Does | Group |
|---|---|---|
| list_mentions | Search and filter the inbox — query text, platform, state, intent, tracker and more. | Find |
| get_mention | Full detail on one mention: classification, engagement, matched terms, any saved quote. | Find |
| list_trackers | Every brand, product and competitor tracker, with its platforms, status and terms. | Find |
| sales_opportunities | The Opportunities tab: purchase-intent and "is there an alternative to X" posts, ranked by priority. | Surface |
| needs_attention | The Needs attention tab: complaints and objections that need a reply before they spread. | Surface |
| analytics_summary | A KPI snapshot for a date range: volume, net sentiment, opportunities, risks, top platforms and themes. | Surface |
| tag_mention | Apply or remove a tag, creating it if the workspace hasn't used it before. | Act |
| update_mention_status | Snooze, resolve, reopen, or mark irrelevant (no undo, same as the inbox's menu). | Act |
| flag_mention | Flag for reply, loop-in, testimonial or ad creative — a note-to-self, not an automated action. | Act |
This list can grow. A connected client calls tools/list to see the current set rather than assuming this table is exhaustive.
Three real jobs — a morning check, a specific question, and cleaning up after handling something — each answered in one exchange instead of a dashboard session.
From MCP/API settings in your workspace, generate an API key. It's scoped to that workspace only, and you can revoke and regenerate it any time — a live secret, not a password you'll reuse elsewhere.
Point your client at the Mentionfront MCP URL with an Authorization: Bearer <key> header. It's the "Streamable HTTP" transport — one POST per call, no SSE stream to keep open — which every current MCP client supports.
Most clients call tools/list automatically on connect. From there it's a normal conversation — the settings page includes a ready-to-paste prompt that also covers the troubleshooting cases below.
Calls are rate-limited per key (120 a minute), which a normal conversation never gets close to — it exists to stop a runaway loop, not to ration ordinary use.
Mentionfront doesn't ship a plugin per agent — it implements the MCP spec once, so anything speaking that spec over Streamable HTTP with a bearer token can connect. That already covers most of what teams use day to day.
Claude.ai, Claude Code and Claude Desktop can all add it as a remote MCP server.
Add it as an MCP server in Cursor's settings to ask about mentions from the editor.
Configured the same way as any other remote MCP server in its config.
Any MCP-compatible agent or custom connector that supports a bearer-token connection.
The tools above are deliberately bounded. That's worth knowing before you connect an agent to real workspace data.
flag_mention is a note-to-self, the same one a teammate leaves in the inbox. Nothing is sent, posted or drafted on your behalf — replies still come from your own accounts.
Marking a mention irrelevant removes it from every view for good, same as the inbox's ellipsis menu. Worth having your agent confirm before it uses that one.
Every id is scoped to the workspace the key belongs to. An id from anywhere else correctly fails instead of returning someone else's data.
| Who | What they used to do | What they ask instead |
|---|---|---|
| Founders | Open the dashboard for a quick gut-check on the brand | "How's sentiment this week, and what's behind it?" |
| Social & CX | Check the Opportunities and Needs attention tabs between other tasks | "Anything urgent since lunch?", then tag and resolve inline |
| Growth | Search the inbox for competitor comparison posts before writing copy | "Pull the last 10 mentions comparing us to Lumaire" |
| Agencies & consultants | Log into each client's workspace separately to pull a status update | A client summary drafted from an analytics snapshot, without switching tabs |