Model Context Protocol (MCP) is Anthropic's open standard for connecting language models to external tools. Arenza exposes its read-only data layer as an MCP server so Claude (Desktop, Code, or any MCP-aware client) can query brand AI visibility data directly.
What the Arenza MCP server can do
- list_brands — return all brands the authenticated user owns, with visibility score and trend.
- get_brand_overview(brand_id) — return AVS, wrong-claim count, top-cited domains for a brand.
- list_prompts(brand_id) — return tracked prompts with last-run rank/mentions/citations.
- get_brand_corrections(brand_id) — list exactly what AI engines get wrong about the brand, with the verbatim AI quote, the correct answer, and the recommended fix.
- score_geo_article(content) — deterministic 0-100 citability score for a draft (no LLM, no auth).
- audit_site(url) — fetch one URL and return a GEO/AEO readiness audit (title, canonical, JSON-LD, content depth).
Run an on-demand scan with run_brand_scan(brand_id) over the PAT (/rpc) transport — it returns a jobId you poll. Fine-grained write tools (mark an opportunity done, add a competitor) are managed in the dashboard today.
Setup for Claude Desktop
- Open Claude Desktop → Settings → Developer → Edit Config.
- Add the Arenza MCP server entry to mcpServers (see config below).
- Restart Claude Desktop.
- In a new chat, type "list my brands" — Claude will call list_brands and render the result.
{
"mcpServers": {
"arenza": {
"url": "https://mcp.arenza.ai/mcp"
}
}
}That's the whole entry — just the URL. On the first call Claude opens a browser to sign in and approve access; it auto-discovers the OAuth endpoints (RFC 8414 + 9728), so there's no token to paste and no header to manage.
Setup for Claude Code (CLI)
Claude Code's CLI accepts MCP servers via the same config schema. Edit ~/.claude/mcp.json:
{
"arenza": {
"url": "https://mcp.arenza.ai/mcp"
}
}Sample queries Claude can answer once connected
- "Which of my client brands lost the most visibility this week?"
- "Show me the top 5 cited domains for Acme Corp."
- "What is ChatGPT getting wrong about Globex right now, and what should we publish?"
- "Which prompts has Acme been mentioned in across ChatGPT and Claude?"
Why MCP-native matters for agencies
Most GEO tools ship a dashboard-only experience. That's fine if your team has 1 brand to look after. If you have 20, you don't want to context-switch between tabs — you want to ask Claude "summarize all client brands' AI visibility for the Monday standup" and have it done in one prompt. MCP makes that possible.
Roadmap
- Q3 2026: fine-grained write tools over MCP (mark opportunity done, schedule digest, add competitor) — OAuth + Dynamic Client Registration is already live.
- Q3 2026: streaming (subscribe to alerts as MCP events).
- Q4 2026: ChatGPT plugin parity (same tool surface).