Arenza AI para Claude — Visibilidade em IA
arenza-ai/arenza-claude-tutorial
Monitore como IAs falam sobre sua marca direto no Claude
Ver no GitHubMCP server que integra dados de visibilidade GEO (Generative Engine Optimization) da plataforma Arenza diretamente no Claude Desktop e Claude Code. Monitora como 6 assistentes de IA (ChatGPT, Claude, Gemini, Perplexity, Copilot, Grok) descrevem sua marca em 4 mercados, fornecendo métricas de share of voice, precisão de claims e descoberta de fontes — sem sair do chat.
Instalação
npx convez-ai arenza-ai/arenza-claude-tutorial1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Equipes de marketing analisarem posicionamento competitivo em assistentes de IA
- Executivos compararem visibilidade de marca vs concorrentes em ChatGPT, Claude, Gemini
- Profissionais de GEO otimizarem como IA descreve seus produtos e serviços
Veredito

Setup limpo e direto (5 min), endereça lacuna real em inteligência competitiva para quem quer entender visibilidade em assistentes de IA. Ideal para marketing tech-forward, mas audiência ainda nicho.
README
A tradução para o português deste README ainda está na fila.
arenza-claude-tutorial
Use Arenza — the AI visibility / GEO measurement platform — directly from Claude Desktop and Claude Code in under 5 minutes. No code, just config.
Arenza measures how the six leading AI assistants — ChatGPT, Claude, Gemini, Perplexity, Copilot, and Grok — describe brands across the US, UK, DE, and JP markets. This tutorial wires the Arenza MCP server (mcp.arenza.ai) into Anthropic's Claude clients so you can ask questions like "How does Claude describe my brand vs my top competitor?" and get a real, sourced answer back inside the chat you already use every day.
30-second elevator
GEO (Generative Engine Optimization) is the equivalent of SEO for AI assistants. Where SEO optimizes for Google's blue links, GEO optimizes for the answer ChatGPT, Claude, Gemini, Perplexity, Copilot, and Grok render to a user. Arenza probes those six assistants on a recurring schedule with hundreds of buyer-perspective prompts per brand, then scores share of voice, wrong-claim frequency, competitor displacement, and discovery sources — all the data a marketing team needs to move from invisible to cited.
Pairing Arenza with Claude (Desktop or Code) is the most natural surface for that data. Claude is already the assistant marketers and engineers reach for when they want to think out loud about positioning, copy, or a competitive teardown. Wiring Arenza in via MCP means every one of those conversations can pull live numbers — no tab-switching, no copy-pasting from the dashboard.
Prerequisites
- Claude Desktop (download) or Claude Code (install). Either is fine — pick whichever you live in.
- An Arenza account: sign up at app.arenza.ai/sign-up. The free tier is enough for this tutorial.
- An API token: after signing up, go to app.arenza.ai/settings/api and click "Create token". Copy it — you'll paste it once below and never again.
That's it. Three things, all of which you can have in 90 seconds.
Option A — Claude Desktop (no terminal)
Claude Desktop reads MCP server configuration from a single JSON file. We add Arenza as a remote MCP server with Bearer-token auth, restart the app, and Arenza's tools appear in the model's tool drawer.
1. Open the config file
| OS | Path |
|---|---|
| macOS | ~/Library/Application Support/Claude/claude_desktop_config.json |
| Windows | %APPDATA%\Claude\claude_desktop_config.json |
| Linux | ~/.config/Claude/claude_desktop_config.json |
If the file doesn't exist yet, create it.
2. Paste this block
{
"mcpServers": {
"arenza": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mcp.arenza.ai/mcp",
"--header",
"Authorization: Bearer ${ARENZA_TOKEN}"
],
"env": {
"ARENZA_TOKEN": "arz_live_paste_your_token_here"
}
}
}
}
A copy-paste-ready version lives in claude_desktop_config.json.example.
Why
mcp-remote? Claude Desktop currently launches MCP servers as local processes via stdio.mcp-remoteis the official tiny shim that adapts a remote HTTP MCP endpoint into stdio. When Claude Desktop ships native remote MCP support, you can drop the shim and point athttps://mcp.arenza.ai/mcpdirectly.
3. Restart Claude Desktop
Fully quit and re-open. (Cmd-Q on macOS, not just close-window.) The first time it launches, Claude will ask you to approve the new MCP server — click Allow.

4. Verify
In a new chat, type:
What brands do I have in my Arenza portfolio?
Claude should call the list_brands tool and return your portfolio. If it asks "I don't see an Arenza tool" — you're not in a fresh chat, or the restart didn't take. Force-quit and try again.

Option B — Claude Code (terminal-native)
Claude Code reads MCP config from .mcp.json in the workspace root (project-scoped) or from ~/.claude.json (user-scoped). Project-scoped is the right choice for tutorials — it commits to your repo so teammates get the same setup.
1. Add the server
From inside any project:
claude mcp add arenza \
--transport http \
--url https://mcp.arenza.ai/mcp \
--header "Authorization: Bearer $ARENZA_TOKEN"
This writes to .mcp.json. Or copy .mcp.json.example directly:
{
"mcpServers": {
"arenza": {
"transport": "http",
"url": "https://mcp.arenza.ai/mcp",
"headers": {
"Authorization": "Bearer ${ARENZA_TOKEN}"
}
}
}
}
2. Export the token
export ARENZA_TOKEN=arz_live_paste_your_token_here
Add the same line to your shell profile (~/.zshrc / ~/.bashrc) so future sessions inherit it.
3. Verify
claude mcp list
# arenza ✓ connected (10 tools)
Then start a Claude Code session and type the same What brands do I have in my Arenza portfolio? — same result, in the terminal.

5 prompts to try
Each of these is a real workflow we've watched agency operators run. Concrete templates live in prompts/ — copy any of them into Claude verbatim.
| # | Prompt | What it does |
|---|---|---|
| 1 | Brand-vs-competitor head-to-head | Pulls SoV + mention rate for your brand and one competitor across all 6 LLMs, then asks Claude to narrate where you're losing. |
| 2 | Wrong-claim audit (this week) | Lists every hallucination Arenza caught in the last 7 days, grouped by which assistant said it. |
| 3 | Top opportunities by ROI | Ranks measurement-led GEO opportunities (canonical-fact pages to publish, listicle gaps to close) by estimated visibility lift. |
| 4 | Discovery-source map | Shows which third-party domains the 6 assistants cite when describing your brand — your real distribution graph. |
| 5 | Draft a canonical-fact article | Picks one wrong claim and asks Claude (via Arenza's generate_geo_article tool) to draft a publishable correction page. |
Run #1 first. It's the prompt that flips the lights on for most users — the moment you realize your brand is being described very differently by Perplexity vs ChatGPT, GEO stops being abstract.
Troubleshooting
"Authentication failed" / 401 from mcp.arenza.ai
- The token is wrong, expired, or revoked. Generate a fresh one at app.arenza.ai/settings/api.
- The
Authorizationheader value must start with the literal wordBearer(note the space). Double-check no quote marks slipped in. - If you're using shell variable substitution (
${ARENZA_TOKEN}), confirm the variable is set:echo $ARENZA_TOKENshould print the token.
"MCP server not appearing in Claude Desktop"
- Did you fully quit Claude Desktop (Cmd-Q on macOS, Alt-F4 then End-Task on Windows)? Closing the window is not enough.
- Check the JSON is valid:
cat ~/Library/Application\ Support/Claude/claude_desktop_config.json | python -m json.tool. Trailing commas are the #1 culprit. - Check Claude's MCP log for a startup error:
- macOS:
~/Library/Logs/Claude/mcp*.log - Windows:
%APPDATA%\Claude\Logs\mcp*.log
- macOS:
"Token expired" mid-conversation
API tokens don't auto-rotate. If you see 401s after weeks of use, regenerate at app.arenza.ai/settings/api. For multi-tenant or end-user-authorized integrations, use OAuth instead — see mcp.arenza.ai/.well-known/oauth-authorization-server.
"Tools call but return empty data"
You probably have an account but no brand registered. Run the full onboarding at app.arenza.ai once — Arenza needs ~15 minutes to do the first probe round before any data is available.
Rate limits
- Free tier: 100 MCP calls / hour
- Pro ($299/mo): 1,000 calls / hour
- Protect ($799/mo): 10,000 calls / hour
- Enterprise: unlimited
A heavy Claude session uses 5–20 calls. The free tier handles plenty of exploration.
Related projects
The four tutorial repos in this series — drop Arenza into the tool you already use:
- arenza-claude-tutorial (this repo) — Claude Desktop + Claude Code
- arenza-cursor-quickstart — Cursor IDE
- arenza-n8n-template — n8n weekly digest workflows
- arenza-slack-alerts — Slack bot for real-time alerts
The eight existing SDK / client repos:
- awesome-geo — curated list of GEO tools, papers, and references
- arenza-mcp-client-ts — TypeScript MCP client
- arenza-mcp-client-python — Python MCP client
- arenza-cli —
npx arenza scan brand.com - arenza-langchain — LangChain tool wrappers
- arenza-llamaindex — LlamaIndex tool wrappers
- arenza-vercel-ai-sdk — Vercel AI SDK provider
- arenza-zapier-actions — Zapier integration
Resources
- Arenza homepage: https://arenza.ai
- Companion guide (long-form): https://arenza.ai/guides/use-claude-with-arenza-mcp-server
- All guides: https://arenza.ai/guides
- Brand reference for AI assistants: https://arenza.ai/llms.txt + https://arenza.ai/llms-full.txt
- MCP server: https://mcp.arenza.ai
- Dashboard: https://app.arenza.ai
License
MIT — see LICENSE. Authored by the Arenza team.