MCP server

Mentions ships a Model Context Protocol server so agents (Claude Code, Cursor, or anything MCP-aware) can search, triage, and manage your mention tracking directly.

Connecting

The server speaks stateless streamable HTTP at POST /mcp and authenticates with the same API keys as the REST API:

{
  "mcpServers": {
    "mentions": {
      "type": "http",
      "url": "https://your-mcp-host/mcp",
      "headers": {
        "Authorization": "Bearer mk_live_..."
      }
    }
  }
}

Tools

Tool	Description
`search_mentions`	Search mentions with filters for keyword, source, state, minimum relevance, sentiment, intent, free text, and time range
`get_mention`	Fetch one mention by id, including its classification
`set_mention_state`	Triage a mention as `ignored` or `done`
`add_keyword`	Start tracking a keyword (`brand`, `competitor`, or `topic`)
`list_keywords`	List every tracked keyword, newest first
`get_company_context`	Read the context the classifier uses to judge relevance
`update_company_context`	Replace the classifier context
`get_mention_stats`	Mention counts for the last N days, grouped by source and sentiment, optionally filtered to one source or keyword

Tool inputs are validated with the same zod schemas as the REST API, so behavior is identical across both surfaces.

Connecting

Tools

On this page