mempalace

A local-first memory palace for AI assistants. Single static binary backed by embedded SQLite (turso). No Python, no ChromaDB, no API keys.

Drop-in replacement for [milla-jovovich/mempalace](https://github.com/milla-jovovich/mempalace) with a ~13MB binary instead of a ~100MB Python environment.

---

The Python version used ChromaDB + SQLite. Under multiple simultaneous MCP clients, SQLite locking caused dropped writes. ChromaDB also carried a large dependency footprint and required Python to be installed.

This reimplementation:

• Ships as a single self-contained binary
• Replaces ChromaDB semantic search with a keyword inverted index (BM25-style scoring via drawer_words)
• Fixes the concurrency problem at the turso layer
• Keeps all 19 MCP tools and all CLI commands fully compatible

Trade-off: Keyword search instead of embedding-based semantic search. Semantic search is deferred until an embedded model is available without network dependencies.

---

git clone https://github.com/bunkerlab-net/mempalace.git
cd mempalace
cargo build --release
# binary is at: target/release/mempalace

Optionally copy to a location on your PATH:

cp target/release/mempalace ~/.local/bin/mempalace

---

claude mcp add mempalace -- /path/to/mempalace mcp

The MCP server runs as a JSON-RPC 2.0 process over stdio. All 19 tools are available immediately after the server starts.

On first use, call mempalace_status — it returns the full memory protocol and AAAK dialect spec in the response, so the AI learns how to use the palace during wake-up.

---

# 1. Initialise a project (creates mempalace.yaml)
mempalace init ~/my-project

# 2. Mine project files into the palace
mempalace mine ~/my-project

# 3. Mine conversation transcripts
mempalace mine ~/my-transcripts --mode convos

# 4. Search
mempalace search "chromadb locking"

# 5. Generate wake-up conte