Attensira MCP¶

Remote MCP server (Streamable HTTP, OAuth 2.1 or workspace API key) - the agent interface for Attensira, an AI-search visibility platform. Instead of switching to a dashboard to read which AI assistants mention a brand, an operator asks the same assistant that holds the numbers: query share of voice, compare windows, find prompts where a competitor appears and you do not, then add the missing prompts in the same conversation. 33 tools across eight groups, 17 read-only.

Server type: Remote (Streamable HTTP)
Auth: OAuth 2.1 (interactive clients) or workspace-scoped static API key (headless callers)
Endpoint: https://mcp.attensira.com/mcp
Health: https://mcp.attensira.com/healthz (unauthenticated reachability check)
Tools: 33 across 8 groups (17 read-only, 3 destructive, 3 credit-spending)
Pricing: Workspace plan (credit-based for automations, prompt additions and ask)
Category: SEO
Built by: Attensira (attensira.com)

Why This Matters for Operators¶

AI-search visibility is the metric every brand now tracks but almost no one acts on in real time. Attensira MCP closes the loop between noticing and changing: an operator editing a pricing page can ask which AI prompts mention a competitor and not the brand, then add three prompts covering the gap without leaving the conversation. The receipts behind every number (answer_history, answer_slice, citation_timeline) mean the visibility data is auditable, which is what makes it safe to act on inside a review cycle.

Tools & Capabilities¶

Group	Tools
Find and measure	`search`, `get_analytics`, `deciding_sources`, `win_plan`, `get_prompt`, `get_page`
Receipts	`answer_history`, `answer_slice`, `citation_timeline`, `page_history`
Track	`add_prompts`, `remove_prompt`, `add_competitor`, `remove_competitor`
Inbox	`list_inbox`, `resolve_inbox_task`, `approve_work_item`, `decline_work_item`
Automations	`list_automations`, `get_automation`, `create_automation`, `run_automation`, `delete_automation`, `update_automation`
Agent	`ask`, `get_session`
Account	`get_account`, `get_brief`, `get_ai_traffic_install`, `connect_github`, `update_settings`
Memory	`record_observation`, `write_skill`

Destructive tools: remove_prompt, remove_competitor, delete_automation. Credit-spending tools: add_prompts, run_automation, ask. Every tool takes an optional project parameter, so one connection reaches every workspace an organization owns (an agency running three domains connects once).

Installation¶

claude mcp add attensira --transport http https://mcp.attensira.com/mcp

Add the URL as a custom connector in Claude, Cursor, VS Code, Codex or ChatGPT and complete the browser OAuth sign-in. For headless callers (n8n, CI, scheduled scripts), mint a static key in Settings, Developer and attach it as the bearer Authorization header; those keys do not expire.

Configuration¶

{
  "mcpServers": {
    "attensira": {
      "url": "https://mcp.attensira.com/mcp"
    }
  }
}

Treat either credential as a workspace credential: an OAuth connection always carries write, and a read-write API key can delete prompts, competitors and automations plus spend credits. A 16-capability REST API is available for code-first integrations with the same key.

Business Relevance¶

Marketing teams pull share-of-voice windows, verify whether deltas are statistically real, and close prompt gaps in one assistant session.
Agencies manage multiple client domains from one connection via the project parameter.
Product teams track which prompts cite the product and which cite competitors before writing release notes.
SEO leads get AI-traffic install data and citation timelines instead of dashboard screenshots.

Integration with CorpusIQ¶

Attensira measures what AI assistants say about a business; CorpusIQ measures what is happening inside it. A CorpusIQ agent preparing a monthly growth report can pull Attensira share-of-voice via get_analytics and combine it with CorpusIQ revenue and funnel data in the same analysis, then use add_prompts to act on the visibility gaps it finds. The citation_timeline receipts give both platforms a common, auditable evidence model.

Limitations¶

New listing: no third-party track record, no public GitHub repo for the server itself.
Write access is broad - destructive and credit-spending tools are one bad prompt away; scope keys read-only where possible.
Credit costs for add_prompts, run_automation and ask are workspace-metered, not published in the docs.
Streamable HTTP only; stdio-only clients need a bridge.