The product as a tool set.

The MCP server lets an agent — Claude, an IDE, anything that speaks the Model Context Protocol — read your reports and manage prompts, topics, tags and brands through typed tools.

Point a client at the endpoint and send the same project key. A generic client config:

{
  "mcpServers": {
    "be-mentioned": {
      "type": "http",
      "url": "https://app.bementioned.ai/mcp",
      "headers": { "X-API-Key": "bm_your_key_here" }
    }
  }
}

Authentication and scope

The key is the scope: no tool takes a project id, and a session token is not accepted — this surface is programmatic-only. A missing or unknown key answers 401 before anything is debited.

Reads require the mcp flag; write tools additionally require mcp_write. The feature groups — actions, shopping, agent analytics, deliverables — also require their own flag, exactly as over REST.

The tool roster

Report tools take the same request bodies as the REST endpoints of the previous chapter, unchanged.

Reports

The report families, one tool per family, plus the single-answer and single-source readers.

get_overview · get_brand_report · get_timeseries · get_matrix · get_rankings · get_prompt_report · get_domain_report · get_url_report · get_chats · get_chat · get_url_content

Listings

The project’s rosters: prompts, topics, tags, brands, the global brand catalog, fan-out queries, engines and markets.

list_prompts · list_topics · list_tags · list_brands · list_global_brands · list_search_queries · list_model_channels · get_markets

Prompt writes

Create, edit, activate, archive and delete prompts as all-or-nothing batches. Activation debits quota once; the daily runs never debit. Requires mcp_write.

create_prompts · update_prompt · activate_prompts · archive_prompts · unarchive_prompts · delete_prompts

Taxonomy and brand writes

Topics, tags, brands, competitor suggestions and the global catalog. Requires mcp_write.

create_topic · update_topic · delete_topic · create_tag · update_tag · delete_tag · create_brand · update_brand · delete_brand · accept_competitor_suggestion · reject_competitor_suggestion · create_global_brand

Classification overrides

Human-tier domain and page-type overrides — the one surface with no REST mirror. Requires mcp_write.

assign_domain_classification · unassign_domain_classification · assign_url_classification · unassign_url_classification

Actions

Read the Actions board and move a card through its states. Requires the actions flag.

get_actions · set_action_state

Shopping

The shopping reports and the product catalog, reads and writes. Requires the shopping flag.

get_shopping_summary · get_shopping_trend · get_shopping_attributes · list_shopping_demand · list_shopping_performance · list_shopping_queries · list_products · list_merchants · list_categories · create_product · update_product · delete_product

Agent analytics

The crawler registry, agent visits and crawlability audits. Requires the agent_analytics flag.

list_bots · get_agent_visits · get_crawlability

Deliverables

The latest version of each deliverable on an Actions card. Requires the deliverables flag.

get_deliverables

Errors and limits

The same 200-per-minute project bucket applies, answering 429 with Retry-After. Inside a tool, failures surface as tool errors of the form status:detail — for example 404:chat not found, or 403:plan_flag_missing:mcp_write.