Tutorials, benchmark analysis, release notes, and deep dives into AI memory systems.
v3.0.0 ships the full MCP server with 19 tools, the temporal knowledge graph (SQLite), specialist agents with per-wing diaries, the PreCompact hook, and the AAAK dialect. Full breakdown of every new feature.
Read More →Within hours of launch, the community caught real problems in the README. Milla Jovovich and Ben Sigman addressed every criticism directly: AAAK token count errors, overstated compression claims, the "+34% palace boost" clarification, and the 100% hybrid score caveats.
Read More →Why verbatim storage beats LLM extraction on recall tasks. A walkthrough of the ChromaDB retrieval pipeline, the 4-layer memory stack (L0–L3), wing/room metadata filtering, and how the +34% retrieval improvement is achieved.
Read More →Step-by-step: export your ChatGPT conversations as JSON, export your Claude projects, run mempalace split to normalise them, and mine them into your palace. Covers the five supported conversation formats.
Read More →Full walkthrough for connecting MemPalace to Cursor, Copilot, and the Gemini CLI via the MCP server. Includes the Gemini CLI auto-setup that handles the MCP server and save hooks automatically.
Read More →Why verbatim storage achieves 96.6% while LLM-extraction tools cap at ~85%. Full feature table covering cost, privacy, MCP support, knowledge graph, enterprise options, and which tool fits which team.
Read More →From daily AI amnesia frustration to 26,900 GitHub stars in 72 hours. The complete origin story — the problem, the architecture decisions, the months of building with Claude Code, the viral launch, and the transparent response to community criticism.
Read More →How to set up code-reviewer, architect, and ops agents in MemPalace. Each agent gets its own wing and AAAK-compressed diary. 50 agents, one line in CLAUDE.md. Comparison with Letta ($20–200/mo).
Read More →