claw-mem: Three-Tier Memory

Three-Tier Memory System for OpenClaw. Store and recall memories across sessions with Episodic, Semantic, and Procedural layers. BM25 + Heuristic retrieval.

Security Scan

The skill's files, scripts, and runtime instructions are coherent with a local three-tier memory plugin: it expects Python and a pip-installable package from the stated GitHub repo, does not request credentials, and does not contain obvious exfiltration or unrelated privileges.

Purpose & Capability

Name/description, SKILL.md, and the three scripts all align: they implement a local memory system, rely on a Python package named 'claw_mem', and operate on a workspace directory. No unrelated credentials, binaries, or system-wide config are requested.

Instruction Scope

Runtime instructions limit actions to installing the package, running the provided scripts, and configuring a workspace path and a search-mode env var. The SKILL.md does not instruct reading unrelated files, harvesting environment secrets, or posting data to third-party endpoints.

Install Mechanism

The README recommends pip installing directly from the GitHub repository (git+https://github.com/opensourceclaw/claw-mem.git). That is a common and expected install path but it does execute code from that remote repository at install-time — review the upstream repository before running pip install.

Credentials

No required environment variables or credentials are declared. Only a search-mode env var (CLAW_MEM_SEARCH_MODE) is suggested, and the scripts use a local workspace path (~/.openclaw/workspace) for SQLite persistence. No disproportionate secret access is requested.

Persistence & Privilege

Skill is not forced always-on and does not request elevated system privileges. It will read/write local SQLite DBs in the specified workspace (episodic.db, semantic.db, procedural.db) which is expected for a memory plugin. Note: the default config in SKILL.md enables autoCapture/autoRecall, which means the plugin will persist conversation data if enabled.

Assessment

This skill appears coherent for a local memory plugin, but take the following precautions before installing: 1) Review the upstream GitHub repo (https://github.com/opensourceclaw/claw-mem) because pip install from a git URL runs code from that repository during install. 2) Install into an isolated environment (virtualenv) to limit impact. 3) Be aware it persists data to ~/.openclaw/workspace (SQLite files) — avoid storing sensitive secrets there and consider its autoCapture/autoRecall settings. 4) If you need stronger assurance, inspect the full claw_mem package source (especially any network code or RPC bridge implementation) before running the bridge. Otherwise the package and included scripts are proportionate to the stated purpose.

claw-mem: Three-Tier Memory System

Local-first memory system with three memory layers and intelligent retrieval.

Prerequisites

pip install git+https://github.com/opensourceclaw/claw-mem.git

Quick Start

Search Memory

python3 {baseDir}/scripts/search.py "project deadlines" --limit 10

Store Memory

python3 {baseDir}/scripts/store.py "User preference" --category preference

Start Bridge (for OpenClaw integration)

python3 {baseDir}/scripts/bridge.py --workspace ~/.openclaw/workspace

Memory Layers

Layer	Purpose	Example
Episodic	Event sequences	Session context
Semantic	Knowledge facts	User preferences
Procedural	Rules	Best practices

Retrieval Modes

Mode	Description	Use Case
keyword	Simple keyword match	Quick lookup
bm25	BM25 ranking	Relevance ranking
hybrid	BM25 + Keyword	Balanced retrieval
enhanced_smart	Full pipeline	Best quality

Set mode via environment:

export CLAW_MEM_SEARCH_MODE=enhanced_smart

Configuration

OpenClaw config (openclaw.config.json):

{
  "plugins": {
    "slots": {
      "memory": "claw-mem"
    },
    "claw-mem": {
      "config": {
        "workspaceDir": "~/.openclaw/workspace",
        "autoRecall": true,
        "autoCapture": true,
        "topK": 10
      }
    }
  }
}

Performance

Operation	Latency
Initialize	~4ms
Store	~8ms
Search	~5ms
Average	~6ms

Advanced

See references for detailed documentation:

• Architecture - Three-tier design
• Retrieval - Search algorithms