memory-tools
Agent-controlled memory plugin for OpenClaw with confidence scoring, decay, and semantic search. The agent decides WHEN to store/retrieve memories — no auto-capture noise. v2 uses file-based storage with optional QMD search (no external APIs required).
Why use this skill?
Enhance your OpenClaw agent with local, persistent memory. Store, retrieve, and manage context using human-readable markdown files without external APIs.
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/gianni-dalerta/memory-toolsWhat This Skill Does
Memory Tools v2 is an agent-controlled persistent memory plugin for OpenClaw that moves away from traditional, noisy, auto-capturing memory systems. Instead of logging every interaction, it follows the AgeMem paradigm, empowering the agent to explicitly choose when to store, update, or retrieve information. By storing data in human-readable markdown files with YAML frontmatter, it ensures transparency and easy debugging. It supports advanced features like confidence scoring, importance weighting, and temporal decay, allowing the agent to distinguish between temporary trivia and critical, long-term instructions. The system is designed to be entirely local, removing the need for external API dependencies or cloud-based vectors.
Installation
- Install via ClawHub:
clawhub install memory-tools - Navigate to the plugin directory:
cd skills/memory-tools - Build the assets:
npm installfollowed bynpm run build - Register the plugin with the gateway:
openclaw plugins install --link .andopenclaw plugins enable memory-tools - Finalize by restarting your gateway:
openclaw gateway restart
Use Cases
This skill is perfect for long-running agentic workflows where context is vital but must remain relevant. Use it to store user preferences that shouldn't be forgotten between sessions, complex project instructions that define agent behavior, or factual snippets discovered during research tasks. Because it includes an optional QMD semantic search layer (BM25 + vector + reranking), it excels at retrieving specific details from large archives of stored markdown memory files without manual indexing.
Example Prompts
- "Store this as a preference: I prefer to receive technical documentation in short, bulleted lists instead of long paragraphs."
- "Search your memory for any instructions regarding the API authentication flow for the current project."
- "Summarize what you know about the user's current project requirements and list the top 3 priorities."
Tips & Limitations
To maximize performance, consider enabling the QMD search layer for high-quality semantic lookups. If you encounter Node ABI mismatches with QMD, you can force basic mode using MEMORY_TOOLS_DISABLE_QMD=true. Remember that since this tool empowers the agent to be the arbiter of its memory, you may occasionally need to guide the agent by explicitly asking it to 'remember' important takeaways. The system defaults to conservative settings regarding legacy migration and instruction injection; verify your openclaw configuration settings to ensure autoInjectInstructions and autoMigrateLegacy are configured to suit your environment.
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-gianni-dalerta-memory-tools": {
"enabled": true,
"auto_update": true
}
}
}Tags(AI)
Flags: file-write, file-read