A memory-intelligent assistant that gives Openclaw the ability to learn from experience by remembering and reusing successful problem-solving traces.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install mia
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 mia 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 MIA (Memory-Intelligent Assistant)?
MIA (Memory-Intelligent Assistant) is a sophisticated learning layer designed to enhance Openclaw Skills by providing a persistent memory of past actions. Instead of treating every request as a brand-new problem, MIA allows the agent to recognize abstract patterns and retrieve previously successful execution trajectories. By synthesizing memory, planning, and feedback modules, it transforms a standard AI agent into a self-improving system that gets faster and more accurate over time.
The core philosophy of MIA centers on universal structural matching and the survival of the fittest. It doesn't just store text; it identifies the underlying logic of a query—such as comparisons, rankings, or temporal references—and ensures that only the most efficient and successful solutions are retained in its long-term memory.
MIA (Memory-Intelligent Assistant) Use Cases
- Reducing latency for complex queries by reusing optimized historical execution plans.
- Enhancing agent reliability in repetitive workflows by following proven successful steps.
- Optimizing multi-step tasks where high-precision planning is required based on past experience.
- Minimizing API consumption by utilizing local planning and cached structural knowledge.
How MIA (Memory-Intelligent Assistant) Works
- The system receives a user query and uses MIA Memory to extract structural features and search for similar historical traces.
- MIA Planner evaluates the retrieved memory to generate an optimized execution plan or creates a new one if no match is found.
- The Openclaw executor performs the necessary actions, such as searching, data verification, and final answer generation.
- MIA Memory analyzes the execution results and stores the new trace only if it exceeds the efficiency score of existing records.
- MIA Feedback collects performance metrics for novel queries to ensure the learning loop remains accurate and helpful.
MIA (Memory-Intelligent Assistant) Setup
To integrate this into your Openclaw Skills environment, follow these installation and configuration steps:
cd /path/to/mia
npm install
Configure your environment variables based on your preferred model provider:
# Example for OpenAI API
export MIA_PLANNER_MODE=api
export MIA_PLANNER_API_KEY=sk-your-key
export MIA_PLANNER_MODEL=gpt-4
# Example for Local Deployment
export MIA_PLANNER_MODE=local
export MIA_PLANNER_URL=http://localhost:8000/v1/chat/completions
export MIA_PLANNER_MODEL=Qwen3-8B
MIA (Memory-Intelligent Assistant) Data Schema & Taxonomy
MIA organizes its persistent data using structured JSONL files to maintain high-speed access and simple portability within Openclaw Skills.
| Data Component | File Path | Description |
|---|---|---|
| Memory Storage | ./memory/memory.jsonl |
Stores execution traces, structural features, and efficiency metrics. |
| Feedback Logs | ./feedback/feedback.jsonl |
Records performance evaluations for newly encountered problem patterns. |
| Structural Features | Internal Schema | Tracks patterns like TIME_REF, DUAL_ENTITY_COMPARE, and RANKING_QUERY. |
MIA (Memory-Intelligent Assistant) Advanced Features
- Survival of the Fittest: Automatically replaces old memory traces with more efficient ones based on step counts and success rates.
- Structural Matching Logic: Uses a weighted similarity algorithm (50% structural, 30% template, 20% semantic) to find relevant memories even with different specific entities.
- Flexible Backend Support: Compatible with local LLMs via Ollama/vLLM and cloud providers like OpenAI and Alibaba DashScope.
- Adaptive Feedback: Intelligently prompts for feedback only on novel problems to avoid user fatigue while maintaining a high-quality knowledge base.
SKILL.md
Loading
Related Openclaw Skills
Flight Tracker
A zero-config aircraft tracking skill for monitoring live flight positions and overhead traffic via OpenSky data.
Proxmox Complete
A comprehensive management tool for Proxmox VE clusters that allows AI agents to control virtual machines, containers, snapshots, and backups.
OpenClaw Config Audit
A robust diagnostic utility that audits OpenClaw environments against 12 production primitives and 8 common setup footguns to prevent data leaks and performance degradation.
Auto Conda Python Env
A smart automation tool that intelligently scans Python project dependencies to create, manage, or reuse matching Conda environments.
VCF Log Insight MCP Explorer
A Model Context Protocol server that empowers AI agents to search and analyze VMware Aria Operations for Logs natively.
QEntity
QEntity provides AI agents with complete financial personhood, allowing them to autonomously earn, hold, and spend money across global payment rails.
Featured*
