A persistent memory system that allows AI agents to store facts, learn from experiences, and track entities across multiple sessions.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install agent-memory
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install agent-memory using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is AgentMemory?
AgentMemory is a robust framework designed to solve the problem of context loss in AI agent workflows. By providing a persistent storage layer, this tool enables agents within the Openclaw Skills ecosystem to maintain continuity over time. It allows for the storage of structured facts, behavioral lessons, and entity-specific metadata, ensuring that every session builds upon the previous one rather than starting from zero.
This capability is essential for building sophisticated, long-running agentic systems that require a deep understanding of historical interactions and project-specific nuances. Using Openclaw Skills like AgentMemory transforms a stateless agent into a knowledgeable assistant that evolves alongside your development process.
AgentMemory Use Cases
- Initializing a new agent session by loading historical context and relevant project facts.
- Capturing durable insights and facts at the conclusion of a conversation to prevent data loss.
- Logging lessons learned after failed attempts or errors to improve future agent performance.
- Building a directory of entities, such as people, roles, or project definitions, to maintain consistency in multi-agent environments.
How AgentMemory Works
- Initialize the memory system using the Python API to establish a connection to the local SQLite database.
- Store information via the remember function for general facts or the learn function for actions and outcomes.
- Query stored data using semantic recall or context-specific filters during active agent workflows.
- Integrate the memory protocol into lifecycle files like AGENTS.md to automate fact extraction and lesson retrieval.
- Persist all data locally in a structured database, allowing for cross-session access and easy backup for Openclaw Skills users.
AgentMemory Setup
To begin using this entry in the Openclaw Skills library, install the package via the CLI:
clawdhub install agent-memory
Then, initialize the memory in your Python scripts:
from src.memory import AgentMemory
mem = AgentMemory()
By default, data is stored at ~/.agent-memory/memory.db, but you can specify a custom path during initialization: AgentMemory(db_path="/path/to/memory.db").
AgentMemory Data Schema & Taxonomy
AgentMemory organizes information into several key segments within a SQLite database:
| Component | Purpose |
|---|---|
| Facts | Stores general information with associated tags for semantic retrieval. |
| Lessons | Tracks actions, contexts, outcomes (positive/negative), and derived insights. |
| Entities | Manages profiles for specific names (people, projects) with flexible metadata JSON. |
| Metadata | Tracks timestamps and source session identifiers for every entry. |
AgentMemory Advanced Features
- Semantic recall capabilities to find information based on query intent rather than exact matches.
- Integration hooks for Clawdbot HEARTBEAT protocols to automate memory updates during session cycles.
- Custom database pathing for self-hosted or shared memory environments.
- Binary outcome tracking (positive/negative) to facilitate reinforcement learning from agent experiences.
- Multi-entity tracking for complex project management within Openclaw Skills.
SKILL.md
Loading
Related Openclaw Skills
eSIMPal API Agent Integration
A comprehensive integration allowing AI agents to browse, purchase, and deliver international eSIM data plans via the eSIMPal API.
ClawClash
ClawClash is a fantasy sports prediction platform built for AI agents to compete in NBA and soccer markets using virtual currency.
Amazon Product Scraper
An automated data extraction tool that uses a built-in browser to scrape Amazon listings and generate comprehensive competitor analysis reports.
Telegram Smart Launcher
A sophisticated UI enhancement that adds dynamic, context-aware CTA buttons to Telegram responses for streamlined interaction.
Last.fm Profile Skill
Integrate Last.fm listening data into your AI workflow to monitor now-playing tracks, top charts, and user profiles.
OpenSubtitles
A read-only skill for searching and downloading subtitles to extract scene context by timestamp for context-aware show analysis.
Featured*
