A persistent memory solution that allows AI agents to store and recall user preferences, project details, and interaction history across multiple sessions.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install tinmem
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 tinmem 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 Tinmem Memory System?
The Tinmem Memory System provides a robust, long-term memory layer designed for AI assistants. As a foundational component of your Openclaw Skills, it ensures that your agent retains context, understands your personal preferences, and remembers specific project decisions long after a session has ended. By utilizing local storage, it offers a secure and fast way to manage agent knowledge without relying on external cloud databases.
This system transforms the AI from a stateless chat interface into a personalized assistant that grows more capable over time. It intelligently categorizes information—from user profiles to complex debugging cases—and injects the most relevant context into the model's prompt window automatically, ensuring highly relevant and personalized interactions.
Tinmem Memory System Use Cases
- Retrieving architectural decisions or technical constraints discussed in previous coding sessions.
- Maintaining a persistent record of user expertise levels and preferred technology stacks.
- Tracking the history of specific bugs and their resolutions across different projects.
- Recognizing and suggesting recurring workflow patterns based on historical usage data.
How Tinmem Memory System Works
- Information Extraction: The system analyzes conversation turns to identify and extract valuable insights automatically.
- Categorized Storage: Data is stored in a local LanceDB database and categorized into specialized schemas like profile, preferences, or entities.
- Semantic Search: When a user asks a question, the agent performs a vector search to find relevant past memories.
- Context Injection: Relevant memories are formatted into agent experience tags and added to the prompt context.
- Continuous Learning: New memories are deduplicated and merged with existing records using LLM-based analysis to maintain a clean knowledge base.
Tinmem Memory System Setup
To begin using the Tinmem Memory System within your environment, ensure you have the necessary package installed as part of your Openclaw Skills.
npm install @openclaw/tinmem
Once installed, the agent will have access to the memory_recall, memory_store, and memory_update tools. No complex database configuration is required as it defaults to a local LanceDB instance.
Tinmem Memory System Data Schema & Taxonomy
Memories are organized into a structured taxonomy to ensure efficient retrieval and logical merging.
| Category | Description | Merge Logic |
|---|---|---|
| profile | User identity, demographics, and roles | Always merge into single profile |
| preferences | Specific likes, dislikes, and recurring habits | Merged based on topic similarity |
| entities | Tools, projects, organizations, and people | Merged when referring to the same entity |
| events | Decisions, milestones, and specific occurrences | Always appended to maintain history |
| cases | Detailed problem-solution pairs and sessions | Always appended for unique case records |
| patterns | Common workflows and methodologies | Merged to refine the pattern description |
Tinmem Memory System Advanced Features
- Automatic memory injection using specialized agent-experience tags for seamless context awareness.
- LLM-powered deduplication to prevent redundant information from cluttering the database.
- Support for detailed levels of retrieval ranging from L0 headlines to L2 full content summaries.
- Importance-based scoring to prioritize the most critical information during context retrieval.
SKILL.md
Loading
Related Openclaw Skills
Project Mode Skill
A sophisticated multi-agent protocol designed to manage complex software development lifecycles through automated task decomposition and parallel execution.
XiaoHongShu Automation Skill V2
A powerful Chrome CDP-based automation solution for full-lifecycle XiaoHongShu account management and content distribution.
Xiaohongshu MCP Patch Pack
A specialized automation toolkit designed to diagnose and repair common deployment issues for the Xiaohongshu Model Context Protocol server.
Investment Meeting Notes Organizer
An AI-powered framework for converting raw investment meeting transcripts into structured, high-fidelity dialogue records.
Smart Image Loader
A versatile utility for AI agents to fetch, manage, and display local or web-based images within their workspace.
Bug Fixing OpenClaw Skill
A comprehensive zero-regression workflow for triaging, reproducing, and fixing bugs within OpenClaw environments.
Featured*
