← Back to Skills Marketplace
845
Downloads
3
Stars
3
Active Installs
5
Versions
Install in OpenClaw
/install mempalace
Description
MemPalace — Local AI memory with 96.6% recall. Semantic search over past conversations, knowledge graph with temporal facts, palace architecture (wings/rooms...
Usage Guidance
This skill appears internally consistent, but before installing you should: 1) confirm the 'mempalace' Python package and its authors (GitHub/PyPI) are trustworthy; 2) install in a controlled environment (virtualenv/container) and inspect the package source if possible; 3) choose and limit the directory you 'mine' so you don't accidentally index sensitive system or personal files; 4) avoid exposing the MCP server to untrusted networks (run it locally and restrict ports); 5) be aware the agent will store verbatim conversation text and diary entries on disk — treat that data as sensitive. If you want higher assurance, request the package's repository/tag/commit hash and a brief changelog before installing.
Capability Analysis
Type: OpenClaw Skill
Name: mempalace
Version: 1.4.0
The mempalace skill provides a local memory system for AI agents, enabling semantic search and knowledge graph management of conversation history. The instructions in SKILL.md are well-aligned with the stated purpose of maintaining long-term context, and the tools provided (e.g., mempalace_search, mempalace_kg_query) are standard for local data retrieval and storage without external network dependencies or suspicious execution patterns.
Capability Tags
Capability Assessment
Purpose & Capability
Name/description, required binaries (mempalace or python3), and the install spec (Python package 'mempalace') all align with a local-memory integration. There are no unrelated credentials or services requested.
Instruction Scope
SKILL.md instructs the agent to query and write verbatim conversation history and a knowledge graph, run a local MCP server (python -m mempalace.mcp_server), and initialize/mine a user-specified directory (~/my-convos). This is coherent with the stated purpose, but mining/indexing could capture any files the user points it at — the instructions rely on the user choosing the correct directory and do not explicitly warn about indexing sensitive system paths.
Install Mechanism
Install spec is a Python package ('mempalace') which matches the README examples (pip install mempalace). This is a standard install pattern, not a direct download from an untrusted URL. Verify the package source (PyPI/GitHub) and maintainers before installing.
Credentials
No environment variables, credentials, or unrelated config paths are requested. The skill requires only local filesystem access to the user's chosen palace directory and the ability to run a local MCP server — these are proportional to a local memory tool.
Persistence & Privilege
The skill writes persistent local data (drawers, diary entries, a knowledge graph) to disk and runs a persistent local service (MCP server) when configured. It does not request always:true or elevated platform-wide privileges, but persistent storage and a local server have privacy implications the user should consider.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install mempalace - After installation, invoke the skill by name or use
/mempalace - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.4.0
- Updated license and attribution: now credits Ben Sigman, Igor Lins e Silva, Milla Jovovich, and adv3nt3.
- No functional or protocol changes; documentation only.
v1.3.0
Version 1.3.0
- Added explicit author attribution in the License & Attribution section ("created by Milla & Ben ([@igorls])").
- No functional changes to the skill code; documentation only.
v1.2.0
- Added a "License & Attribution" section with full MIT license text and attribution to MemPalace contributors.
- No changes to protocol, tool descriptions, or usage instructions.
v1.1.0
- Updated to version 1.1.0.
- Description clarified with product branding ("MemPalace — Local AI memory...").
- No functional or protocol/tool changes—documentation and presentation improvements only.
- No user-facing commands, setup, or usage instructions were changed.
v1.0.0
- Initial release of mempalace: a local AI memory system with 96.6% recall and top LongMemEval performance.
- Semantic search and powerful filtering for fast context recall from conversation history.
- Temporal knowledge graph tracks facts, entities, and their changes over time.
- Organizes memories hierarchically as wings (e.g., people, projects), rooms, and drawers.
- All features run locally: no cloud, no API keys, full user control and privacy.
- Includes session diary, palace graph traversal, and advanced fact management tools.
Metadata
Frequently Asked Questions
What is mempalace?
MemPalace — Local AI memory with 96.6% recall. Semantic search over past conversations, knowledge graph with temporal facts, palace architecture (wings/rooms... It is an AI Agent Skill for Claude Code / OpenClaw, with 845 downloads so far.
How do I install mempalace?
Run "/install mempalace" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is mempalace free?
Yes, mempalace is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does mempalace support?
mempalace is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created mempalace?
It is built and maintained by boluobobo (@wanikua); the current version is v1.4.0.
More Skills