Support & FAQ

Get help

• Email: support@marcopolo.dev
• Slack Community: [Join the MarcoPolo community]() . Ask questions, share workflows, and see what others are building.

Troubleshooting

"My AI can't connect to MarcoPolo"

Make sure the MCP server URL is exactly https://mcp.marcopolo.dev. Check your AI tool's MCP configuration. Each tool has a slightly different setup. See the integration guide for your specific tool.

"My data source isn't showing up"

After adding a data source, ask your AI: "List my data sources." If it doesn't appear, check the connection status in the web UI. The most common issue is a firewall blocking MarcoPolo's IP. Make sure 34.208.3.240/32 is allowlisted.

"My query is failing or returning errors"

Ask your AI to show you the exact error message. Common causes: wrong table or column names (ask your AI to run get_schema first), syntax issues (MarcoPolo provides syntax guides for each data source, but complex queries may need manual adjustment), and permission issues on the data source side.

"The AI is giving wrong results"

This is almost always a context problem. Write a RULES.md for your data source with metric definitions, canonical tables, and known gotchas. The 15 minutes you spend writing RULES.md will pay for itself many times over.

"My workspace seems slow to start"

Workspaces scale down when idle and spin back up on the next request. The first request after a period of inactivity may take a few seconds longer while the container starts. Subsequent requests are fast.

"The Codex plugin isn't loading"

Make sure the plugin was cloned to ~/.codex/plugins/marcopolo and that ~/.agents/plugins/marketplace.json references the correct path. After any changes, restart Codex. Verify the plugin appears in your marketplace and that the skills query-workflow, using-marcopolo, and workspace-navigation are available. If skills don't load, check that the marketplace.json format matches the example in the Codex plugin guide.

Limits and constraints

For higher limits or enterprise configurations, contact support@marcopolo.dev.

Frequently asked questions

How does authentication work? There are two separate things: connecting to MarcoPolo and connecting data sources. Your MarcoPolo account is created the first time you sign in with Google, GitHub, Microsoft, or email. Users from the same email domain are grouped into the same company. Data source connections are separate. Each one has its own setup flow (OAuth, username/password, API keys, etc.) and MarcoPolo walks you through the right one.

Is my data secure? Yes. Credentials are encrypted and never exposed to the AI. Each user gets an isolated workspace. Data passes through MarcoPolo but isn't stored or replicated. See Security for the full model.

Can the AI see my credentials? No. The AI operates in user mode. A separate privileged layer injects credentials and executes queries. The AI never has access to passwords, tokens, or connection strings. See Credential Isolation.

Is my data used for AI training? No. Customer data never enters any training pipeline.

Can I see what the AI did? Yes. Everything is in your conversation history and as files in your workspace. Every query, script, and artifact the AI creates is saved. Browse them in the web app or ask your AI to list them.

Can the AI or MarcoPolo access data I didn't ask for? No. The AI only sees query results you explicitly request. It has no background access to your systems. MarcoPolo infrastructure does not access your workspace contents.

Can I share a data source outside my company? No. Sharing is restricted to users on the same email domain.

What happens when I'm inactive? The workspace runtime (compute) scales down. Your files, scripts, cached data, and artifacts persist. When you make a new request, the workspace spins back up and everything is where you left it.

Can I connect to a local database? MarcoPolo connects to data sources reachable over the public internet. If your database runs only on your laptop, you'll need to expose it via a tool like ngrok. For private network connectivity, see enterprise options.

What AI assistants are supported? Claude (Web, Desktop, Code), Codex, ChatGPT, Cursor, VS Code / Copilot, Replit, and any tool that supports MCP. See Getting Started for setup guides.

How do I add a new data source? Ask your AI ("I want to add a PostgreSQL database") or go directly to mcp.marcopolo.dev/app/datasources. See Data Sources for details.

What's the IP address for allowlisting? 34.208.3.240/32. Add this to your database firewall or network access list.