skills.rest
yoloshii
yoloshiiCOMMUNITY
Explore all skills by this author →
GitHub

clawmem

Provide local embedded search and retrieval for MCP-compatible AI agents.

197|32|Updated Feb 6, 2026
Search Tags:#AI agent#contextual retrieval#document indexing#memory layer#MCP client#retrieval-augmented search#local search engine
One-click install
npx skills add https://github.com/yoloshii/ClawMem --skill clawmem
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: clawmem
Source: https://github.com/yoloshii/ClawMem/tree/main
Command: npx skills add https://github.com/yoloshii/ClawMem --skill clawmem

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires node-llama-cpp, bun, sqlite, python, numpy, pandas, sqlalchemy, and includes scripts (resource) and references (resource) and assets (resource) components.

What problem does it solve?

This Skill solves the problem of providing AI agents with access to a persistent memory layer that allows them to leverage past decisions, preferences, and information to enhance context-aware responses.

Core Features & Use Cases

  • Contextual Retrieval: Automatically surfaces relevant context on every prompt for AI agents.
  • Session Bootstrap: Boots up sessions with user profiles, latest handoffs, recent decisions, and stale notes.
  • Decision Capture: Captures decisions, preferences, milestones, and problems from session transcripts.
  • Import and Synthesize: Imports conversation exports from various sources and synthesizes structured information.
  • Handoff Generation: Generates handoffs at session end to ensure continuity between sessions.
  • Memory Learning: Learns what matters through a feedback loop that boosts referenced notes and decays unused ones.
  • Query Intent Classification: Classifies query intent to guide search strategies.
  • Multi-Graph Traversal: Traverses semantic, temporal, and causal graphs for in-depth analysis.
  • Memory Evolution: Evolves memory metadata as new documents create or refine connections.
  • Causal Relationship Inference: Infers causal relationships between facts.
  • Contradiction Detection: Detects contradictions between new and prior decisions.
  • Cross-Entity Merge Prevention: Prevents context bleed in derived insights.
  • Structured Triple Injection: Injects knowledge-graph facts as structured triples.
  • Focus Topic Boost: Applies a focus topic boost for session-specific contexts.
  • Document Quality Scoring: Scores document quality using structure, keywords, and metadata richness signals.
  • Co-accessed Document Boost: Boosts co-accessed documents.
  • Query Decomposition: Decomposes complex queries into typed retrieval clauses.
  • Stale Embedding Cleanup: Cleans stale embeddings before embed runs.
  • Transaction-safe Indexing: Ensures transaction-safe indexing.
  • Deduplicated Hook Observations: Deduplicates hook-generated observations.
  • Temporal Navigation: Navigates temporal neighborhoods around any document.
  • Frequently-REVISED Memory Boost: Boosts frequently-revised memories.
  • Lifecycle Management: Manages document lifecycle through pin/snooze and archival policies.
  • Auto-routing Queries: Auto-routes queries via memory_retrieve.
  • Project Issue Syncing: Syncs project issues from Beads issue trackers.
  • Heavy Maintenance Lane: Provides a quiet-window heavy maintenance lane for background tasks.

Quick Start

Use the clawmem skill to initialize a new memory vault at ~/notes, name it notes, and index your documents.

clawmem bootstrap ~/notes --name notes

Frequently Asked Questions about clawmem

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I add a persistent memory layer to an MCP client for contextual retrieval?▼

To add a persistent memory layer to an MCP client for contextual retrieval, use a local embedded search engine that automatically surfaces relevant context on every prompt and stores past decisions in a local SQLite vault.

How does multi-graph traversal and intent classification improve AI agent memory?▼

Multi-graph traversal improves AI agent memory by navigating semantic, temporal, and causal relationships, while intent classification guides search strategies and decomposes complex queries into typed retrieval clauses for precise context.

Can I use local SQLite and Python environments for AI agent session bootstrapping?▼

Yes, you can use a local SQLite vault and Python environment with libraries like numpy and pandas for AI agent session bootstrapping, which initializes sessions with user profiles, recent decisions, and handoffs.

What is the best way to detect contradictions and infer causal relationships in AI memory?▼

The best way to detect contradictions and infer causal relationships in AI memory is using an advanced retrieval engine that evaluates new facts against prior decisions and injects structured knowledge-graph triples.

Does clawmem require a dedicated GPU and node-llama-cpp for document indexing?▼

clawmem requires node-llama-cpp, bun, and sqlite for transaction-safe document indexing, while an optional GPU provides dedicated performance for processing local search engines and embedding runs.

How do I manage memory lifecycle and clean stale embeddings in a local search engine?▼

You manage memory lifecycle and clean stale embeddings in a local search engine by applying pin/snooze archival policies, boosting frequently-revised memories, and executing heavy maintenance during quiet background windows.

Related Skills

Explore More Software Engineering Skills→
x-TheFoxx-TheFox

llama-cpp

Run local GGUF inference across CPU, Apple Silicon, and GPU backends.

Community
Intermediate
blackplume233blackplume233
12

conversation-management

Manage multi-turn dialogue context and session lifecycle for AI conversations.

Community
Intermediate
ChrisWilesChrisWiles

wiki-history-ingest

Route conversation history from Claude, Codex, Hermes, and OpenClaw to ingestion skills.

Community
Advanced
BoraPerusicBoraPerusic

rag-engineer

Design and optimize RAG pipelines with embeddings, vector databases, and chunking strategies.

Community
Advanced