Asva AIPower the future
Model Context Protocol · Updated September 2026

The only AI-visibility platform you can install as an MCP server.

Connect Asva AI to Claude, ChatGPT, Cursor or VS Code and ask about your AI-search visibility, citations and share of voice in the thread you already work in — backed by your live workspace, not a training-data guess.

Endpointhttps://mcp.asvaai.com/mcp

What the Asva AI MCP server is

An MCP server is a standard endpoint that lets an AI assistant call external tools.

Asva AI's exposes your answer-engine-optimization tracker as those tools. The Asva AI MCP server exposes your live tracker — mention rate, share of voice, citations, sentiment and competitor rank across ChatGPT, Perplexity, Gemini, Copilot, Grok, Meta AI and Google AI Mode — as OAuth-authed MCP tools any compatible assistant can call.

Once you add it, you can ask "what's my share of voice on Perplexity this month?" and the model calls Asva, reads your workspace, and answers in chat. No dashboard round trip, no export, no copy-paste. The dashboard is where you build; the MCP server is where you ask.

It authenticates over OAuth 2.1 with PKCE and speaks Streamable HTTP (MCP spec 2025-03-26), so a hosted client connects with a sign-in pop-up and no token handling.

Why it is the only one of its kind

No competing AI-visibility platform ships an installable MCP server for AI assistants today.

Only-of-its-kind

Every comparable tool is a dashboard plus, at best, a REST API you write code against. None exposes its tracker as an MCP server you drop into an assistant. This is a specific, falsifiable claim — verified across the category in September 2026.

In-context, not another tab

Your visibility data reaches the assistant you already brief, prompt and write in. Ask the number in the same thread where you draft the fix — no dashboard round trip.

Composable

Because it is MCP, your data joins whatever else the client can reach — your repo in Cursor or VS Code, your other connectors in Claude — so an agent can read your share of voice and act on it in one turn.

The tools the MCP server exposes

The server exposes your live AEO tracker as read-only tools any assistant can call.

Every analytics tool takes a parent_brand_id, so list_brands is always the first call — resolve the name to its id, then query. All tools are read-only except the two writers (the Quick Report audit and the llms.txt generator). The full contract for each tool lives in the reference.

Entry point

list_brands

Resolve a brand name to the id every other tool needs.

Headline & competitive analytics

get_brand_visibility, get_share_of_voice, get_competitors, get_leaderboard, get_platform_breakdown, get_sentiment_analysis, get_regional_visibility

Mention rate, share of voice, competitor rank, per-platform and per-region rollups.

Citations & shopping

get_citation_metrics, get_gpt_shopping_visibility

Top citing domains and category mix; visibility inside ChatGPT Shopping cards.

Prompts & raw responses

list_topics, list_prompts, get_prompt_responses

Drill from a tracked topic to the prompts and the actual AI answers, with citation URLs.

Quick Report audit (write)

generate_visibility_audit, list_my_visibility_reports, get_visibility_report

Kick off a self-serve audit on any domain (runs ~20–30 min), poll, then fetch the full report.

Content (write)

generate_llms_txt

Generate an llms.txt for a site — it pairs naturally with the AEO workflow.

Connect it to your assistant

Add the endpoint once, and your assistant queries your visibility data in-context.

Claude Web & Desktop

Settings → Connectors → Add custom connector → paste the URL → sign in when the OAuth pop-up appears.

Full steps in the docs

Claude Code (CLI)

Run claude mcp add --transport http asvaai https://mcp.asvaai.com/mcp, then confirm with claude mcp list.

Full steps in the docs

Cursor

Add a server entry to ~/.cursor/mcp.json (or the per-project file); Cursor opens a browser to finish OAuth.

Full steps in the docs

VS Code (Copilot agent mode)

Add an HTTP server to .vscode/mcp.json with the top-level servers key, click Start, then pick it in the Tools picker.

Full steps in the docs

Windsurf

Settings → Cascade → MCP Servers, add the config with serverUrl, save and restart.

Full steps in the docs

Any other MCP client

Use the HTTP transport URL directly, or bridge with npx mcp-remote. Spec-compliant clients self-register via Dynamic Client Registration.

Full steps in the docs

Connect Asva AI to ChatGPT

ChatGPT reaches your Asva workspace through a custom connector you enable in settings.

Custom Connector / Developer mode

  1. Open Settings → Connectors. On lower tiers, enable Developer mode under Advanced settings first.
  2. Click Create, name it Asva AI, and paste the URL https://mcp.asvaai.com/mcp.
  3. Set Authentication to OAuth and sign in with your Asva AI account.
  4. Enable the connector for the chat, then prefix your message with "Use Asva AI to…" so the model reaches for the tools.

ChatGPT app (Apps SDK) — on the roadmap

A published ChatGPT app built on the Apps SDK is planned, not shipped. Today you connect Asva AI as the custom connector above; there is no published Asva AI app in the ChatGPT directory yet. We will update this page and the docs when that path is live.

What you can ask

Ask in plain language; the model picks the right tool and answers.

Headline & trend

  • How is our brand doing on AI search this month?
  • What's our share of voice on Perplexity vs Gemini in the US, last 30 days?
  • Break our visibility down by platform — where are we weakest?

Competitive

  • Show the competitor matrix and who's beating us on ChatGPT.
  • Which topics do we rank worst on versus competitors?

Citations & sentiment

  • Which domains cite us most in AI answers, and what type are they?
  • Why is our sentiment low this week? Show example snippets.
  • How visible are our products inside ChatGPT Shopping?

Drill-down & workflow

  • List the prompts where we're not mentioned for topic X.
  • Run a visibility audit for stripe.com, then summarise it.
  • Draft this week's AI-visibility digest from the last 7 days.

All numbers come from your live Asva workspace; the assistant cannot answer these from its training data — that is the point.

Security and access

Every call is a short-lived OAuth bearer, scoped to brands you can access.

  • Each tool call is authenticated with a short-lived OAuth 2.1 bearer token; the assistant never sees your password.
  • Hosted clients authenticate through an OAuth pop-up — you never paste a token. Self-hosted bridges use a token you generate in the dashboard.
  • A call only returns data for brands the authenticated user can access. Cross-account reads are rejected at the backend layer.
  • Every call is audit-logged with your user id and timestamp; the server stores no credentials.
  • Disconnect at any time from the Connectors page in your dashboard — that revokes every token tied to that client.

For the full data-handling policy, see the Privacy Policy and the Terms of Service.

Frequently asked questions

Common questions about connecting Asva AI as an MCP server.

What is the Asva AI MCP server?+

A Model Context Protocol endpoint that exposes your live AI-visibility data as tools any MCP client can call. Once connected, an assistant reads your Asva AI workspace and answers questions about mention rate, share of voice, citations, sentiment and competitor rank in natural language.

Which assistants can connect?+

Claude (Web, Desktop and Code), ChatGPT, Cursor, VS Code (Copilot agent mode), Windsurf, and any MCP-compatible client via the HTTP transport. The full setup for each is in the docs at /docs/mcp.

Do I need an API key?+

No. Hosted clients authenticate through an OAuth pop-up, so you never paste a token. Self-hosted bridges use a token you generate in the dashboard.

Is my data safe?+

Every call rides a short-lived OAuth 2.1 bearer, scoped to the brands you can access, audit-logged, and revocable from the Connectors page in your dashboard. The server stores no credentials.

Does any competitor offer this?+

No AI-visibility platform ships an installable MCP server today. Every comparable tool is a dashboard plus, at best, a REST API you write code against. The MCP server is unique to Asva AI as of September 2026.

Is this the same as the agentic-commerce MCP?+

No. This server reads your AI-visibility data. The separate agentic-commerce MCP (a different product on a different domain) lets AI agents transact against a product catalog. Different servers, different jobs.

Put your AI visibility in the thread you work in

Connect the endpoint to your assistant with the step-by-step docs, or run a free scan to see where your brand stands in AI search first.