Skip to main content

Troubleshooting

MCP server not responding

Symptom: Your AI tool can’t connect to BaseLayer, or tool calls fail. Fix:
  1. Make sure the BaseLayer desktop app is running (check for the menu bar icon)
  2. Restart the desktop app
  3. If using remote MCP (Claude.ai, ChatGPT): check your internet connection
  4. If using local MCP (Claude Code, Gemini CLI): verify the server is on port 44044
# Check if the local MCP server is running
curl http://localhost:44044/health

Chrome extension not capturing

Symptom: Conversations aren’t appearing in your vault. Fix:
  1. Click the BaseLayer extension icon and check for the green status dot
  2. If the dot is gray: you’re not on a supported site
  3. If the dot is red: try signing out and back in
  4. Refresh the AI page and try again
  5. Make sure the desktop app is running

”0 entities” in vault

Symptom: You’ve been using BaseLayer but your vault shows zero entities. Fix: This usually means an encryption key issue.
  1. Restart the desktop app
  2. You might need to re-enter your passphrase
  3. If it still doesn’t work, check Settings → Vault for error messages
If you’ve forgotten your passphrase, we can’t decrypt your vault. We don’t have access to your passphrase.

Dream engine not processing

Symptom: Conversations are captured but no entities are being extracted. Fix:
  1. Open the desktop app → Settings → Dream Engine
  2. Check the queue; there may be a backlog
  3. Click Trigger Dream to manually start processing
  4. If the queue is stuck, restart the desktop app

MCP tools not appearing in Claude.ai

Symptom: You connected BaseLayer but Claude doesn’t use the tools. Fix:
  1. Go to Settings → Integrations and verify BaseLayer is listed
  2. Try removing and re-adding the integration
  3. Start a new conversation (tools may not appear in existing ones)

Slow search results

Symptom: recall takes a long time to return results. Fix: Large vaults (10,000+ entities) may take longer to search.
  • Use entity_type to narrow your search
  • Use depth: "shallow" for faster results
  • Reduce the limit parameter