Troubleshooting

Cause: The server was added without --transport http. Claude Code defaults to stdio, so it treats the URL as a local command to execute instead of an HTTP endpoint to call.

How to fix: remove it and re-add with the transport flag, then sign in:

claude mcp remove prowlo
claude mcp add --transport http -s user prowlo https://api.prowlo.com/mcp
claude mcp login prowlo

Full walkthrough in the Claude Code guide.

Cause: A Claude plan limit, not a Prowlo problem. Neither produces a useful error — the option is just missing or does nothing.

• Free — One custom connector, total. If you already have one connected, Claude will not let you add Prowlo until you remove it.
• Pro & Max — Multiple connectors. Add yours under Customize → Connectors → + → Add custom connector.
• Team & Enterprise — Only an organization Owner can add a connector, under Organization settings → Connectors → Add → Custom → Web. Everyone else then enables it under Customize → Connectors — a member cannot add it themselves.

How to fix: Blocked by either of these? Claude Code, VS Code, Zed, and Cursor connect to Prowlo per-user over the same hosted endpoint — no org permission, no connector limit, and no waiting on an Owner.

Cause: Your configuration file may not be loaded correctly, or there's a JSON syntax error.

How to fix:

• Restart your AI tool completely (quit and reopen the application)
• Validate your JSON configuration file — check for missing commas, unclosed brackets, or typos
• Ensure "prowlo" is nested inside the right top-level object, not at the root — and note the key differs per client: mcpServers (Claude Desktop, Cursor, Cline, Windsurf), servers (VS Code), context_servers (Zed). A config copied between tools will be silently ignored.
• Verify the configuration file location matches what your tool expects (e.g., ~/.config/cursor/mcp.json for Cursor)

Cause: Your API key is incorrect, expired, or has been revoked.

How to fix:

• Verify the key starts with prowlo_
• Check your Prowlo dashboard — go to API Access to confirm the key is still active
• Make sure there are no extra spaces, line breaks, or special characters around the key in your configuration
• Verify your Authorization header is formatted exactly as: Bearer prowlo_your_key (with exactly one space after "Bearer")
• If the key is revoked, generate a new one in your API Keys settings

Cause: Network connectivity issue or the Prowlo API may be temporarily unavailable.

How to fix:

• Check your internet connection — try visiting another website to confirm connectivity
• Verify you're using the correct endpoint: https://api.prowlo.com/mcp
• Test the API health by visiting https://api.prowlo.com/health — you should see {"status":"ok"}
• If you're behind a corporate proxy or firewall, ensure outbound HTTPS traffic to api.prowlo.com is allowed
• Wait a few minutes and retry — temporary outages are rare but can happen

Cause: You've made more than 60 requests in the last minute with your API key.

How to fix:

• Wait 60 seconds for the rate limit window to reset
• Check the X-RateLimit-Reset header in the response for the exact reset time
• Consider reducing the frequency of your requests or implementing exponential backoff in your integration
• If you consistently hit rate limits, contact us — we can discuss higher limits for your use case

Cause: Your MCP session timed out (sessions expire after 2 hours) or the server restarted.

How to fix:

• This is expected behavior — your AI tool will automatically reconnect and create a fresh session
• No action is needed; just retry your request
• If the error persists, restart your AI tool completely to force a reconnection

Cause: No records match your search, or your Watchers haven't crawled any sources into your Dataset yet.

How to fix:

• Log into your Prowlo dashboard and check whether records exist in your Dataset
• Verify you have at least one active Watcher — with no Watchers, Prowlo crawls nothing
• Try a broader query in search_dataset — semantic search matches on meaning, so widen the topic or drop restrictive filters
• Use list_records with no filters to see the most recent records as raw data
• Wait 24–48 hours after creating a Watcher — our system needs time to crawl, clean, and embed records into your Dataset