No Built-in LLM: Local CPU/GPU execution is not supported. You must connect to an external self-hosted LLM or a cloud-based LLM provider.
Embedding Model Limitations
Built-in Embedder: Large document uploads on the default built-in embedder will exhaust CPU/RAM resources and cause a 502 error (process exit).
Mitigation: Use an external embedding provider for large documents, or upgrade from the Starter tier to the Professional tier for higher compute limits.
Workspace Search Settings
Accuracy Optimized Search: High CPU/RAM consumption. Using this mode on workspaces with large file counts can cause the instance to freeze or become unresponsive.
Mitigation: Set the workspace search mode to Default in the Workspace settings.
Feature Restrictions
Custom Agents: Disabled due to security policies restricting arbitrary code execution. Returns system administrator error.
MCP (Model-Context Protocol): Unsupported due to security policies restricting arbitrary code execution.
Workaround: For Custom Agents or MCP support, use AnythingLLM Desktop or a self-hosted Docker instance.