Skip to main content

​
Knowledge Base (MemPalace)

The Knowledge Base page provides semantic search over your code, documentation, and project knowledge. It is powered by MemPalace, an open-source local memory system that indexes content into a searchable vector database.
GitHub: github.com/milla-jovovich/mempalace — MIT License, 27k+ stars
Everything runs locally — no external APIs, no data leaves your machine. Knowledge Base

​
Enabling MemPalace

MemPalace is an optional feature. When you first visit the Knowledge Base page, you’ll see an “Enable MemPalace” button.
  1. Go to Dashboard > Knowledge
  2. Click Install MemPalace
  3. Wait for the installation to complete (installs via pip)
Requirements: Python 3.9+ The installation adds the mempalace package and its dependency chromadb (the vector database engine).

​
How It Works

MemPalace organizes knowledge using a “memory palace” metaphor: When you index a directory, MemPalace:
  1. Scans all files (respecting .gitignore)
  2. Splits content into chunks of ~800 characters
  3. Classifies each chunk into a room based on content and file path
  4. Stores vector embeddings in ChromaDB for semantic search
All data is stored locally at dashboard/data/mempalace/.

​
Adding Sources

Sources are directories you want to index — project code, documentation folders, etc.
  1. Go to the Sources tab
  2. Enter the directory path (absolute path on your machine)
  3. Optionally set a label (display name) and wing override (custom project name)
  4. Click Add
MemPalace only re-processes files that have changed since the last indexing, so re-indexing is fast.

​
Indexing (Mining)

After adding sources, click the Play button next to a source to index it, or use Index All to process everything. Mining runs in the background — you can navigate away and check progress on the Status tab. The status indicator shows when mining is active.

​
Searching

The Search tab provides semantic search across all indexed content:
  1. Type your query in natural language (e.g., “how does authentication work”)
  2. Optionally filter by wing (project) or room (category)
  3. Press Enter or click Search
Results show:
  • The matched content snippet
  • Source file path
  • Similarity score (higher = more relevant)
  • Wing and room tags
Semantic search finds results by meaning, not just keyword matching. For example, searching “auth” will also find content about “login”, “session tokens”, and “JWT”.

​
Permissions

​
API Endpoints

All endpoints require authentication and appropriate permissions.

​
MCP Integration

For advanced users, MemPalace can also be registered as an MCP server for direct Claude Code access:
This enables Claude to query your knowledge base directly during conversations.