liza-studio/skillmem

Overview

Skillmem is a local memory layer for coding agents that learns, recalls, reinforces, and decays skills over time. It stores memories in SQLite with provenance tracking and requires no API key or cloud service. Access is controlled by explicit trust grants from the user.

Best for

Best for
Developers building self-improving coding agents that need persistent, local, trust-controlled skill memory.

Use cases

• Persist coding skills across agent sessions
• Recall previously learned solutions with provenance
• Reinforce or decay memory strength based on usage

How to use

Tools exposed

• mem_search
• mem_get
• mem_list
• mem_write
• mem_update
• mem_learn
• mem_recall
• mem_reinforce
• mem_pin
• single-session-assistant
• knowledge-update
• single-session-user
• multi-session
• single-session-preference
• temporal-reasoning

Tested with

Claude Desktop, Claude Code, Cursor, Windsurf

Notes

Skillmem is a local memory layer for coding agents that learns, recalls, reinforces, and decays skills over time. It stores memories in SQLite with provenance tracking and requires no API key or cloud service. Access is controlled by explicit trust grants from the user.

4 stars on GitHub. Last updated 2026-09-16. Licensed Apache-2.0.

Use cases

• Persist coding skills across agent sessions
• Recall previously learned solutions with provenance
• Reinforce or decay memory strength based on usage

Pros

• Fully local with SQLite, no external dependencies
• Provenance on every memory improves traceability
• Trust model gives the user explicit control

Cons

• Requires Python integration, not language-agnostic
• Decay and reinforcement behavior may need tuning
• Limited to what the agent explicitly learns and stores

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.