engram

What This Skill Does

Engram is a robust, local semantic memory system designed for OpenClaw AI agents. Unlike volatile context windows, Engram provides persistent, searchable storage that mimics biological memory through a decay mechanism. It allows your agent to store facts, decisions, preferences, and events, ensuring that context remains available across sessions without needing external databases or API keys. By utilizing semantic embeddings, Engram enables agents to retrieve relevant information based on conceptual similarity rather than simple keyword matching. It features automatic relationship mapping, which allows the agent to understand how different pieces of information, such as decisions and their supporting facts, are interconnected.

Installation

To integrate Engram into your OpenClaw environment, execute the following command in your terminal: clawhub install openclaw/skills/skills/dannydvm/engram-memory Once installed, verify the installation by running engram stats to check your memory database status.

Use Cases

Engram is ideal for long-running agent projects. Use it to keep track of technical project requirements, maintain client-specific preferences (like coding styles or communication tone), archive critical decisions to avoid repeating past mistakes, and manage complex event logs that inform future agent reasoning. It is particularly effective for developers building agents that need to act as a personal research assistant or a long-term project manager.

Example Prompts

1. "Engram, recall what the client decided regarding the database migration last month so we can proceed with the current schema update."
2. "Store this new fact in Engram: The primary color palette for the dashboard project is #0047AB, and tag it as 'branding'."
3. "Search for all previous decisions tagged with 'ads' to see if we have already evaluated the impact of our current budget constraints."

Tips & Limitations

• Decay Matters: Engram uses a biological decay algorithm. Memories that are not accessed regularly will lose salience. Make sure to query important data occasionally to keep it active.
• Keep It Typed: Always categorize your memories using the available types (fact, decision, preference, etc.) to improve search precision.
• Manual Curation: While auto-deduplication handles 92% of overlaps, periodically run engram export to audit your data and perform manual house-cleaning.
• Limitations: Since Engram is purely local, it does not sync across different machines automatically. Ensure you use the export and import functions if you are switching hardware.