A high-performance Model Context Protocol server that injects switchable expert memory and dynamic workflows into AI agents.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install engram-mcp-server
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 engram-mcp-server 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 Engram MCP Server?
Engram MCP Server is a specialized tool designed to provide AI coding agents with a structured and persistent memory framework. By moving beyond simple knowledge retrieval, it encapsulates expert memory into loadable resources that define who the agent is, what it knows, and how it should think. This makes it an essential addition to the Openclaw Skills collection for developers who need their AI assistants to maintain context across sessions.
The server utilizes a zero-vector-dependency architecture, meaning it operates efficiently without requiring heavy external databases like ChromaDB. It employs an index-driven layered loading system where core components like roles and workflows are permanently resident in the context, while specific knowledge fragments and examples are loaded on demand to optimize performance and token usage.
Engram MCP Server Use Cases
- Rapidly switching between different expert personas (e.g., Backend Architect vs. DevOps Engineer) in a single session.
- Persisting user preferences, project-specific facts, and technical constraints across multiple AI conversations.
- Managing complex domain knowledge and few-shot examples within MCP-compatible clients like Claude and Cursor.
- Maintaining a dynamic memory of tool execution trajectories to improve future AI decision-making.
How Engram MCP Server Works
- The user adds the Engram server to their MCP-compatible IDE or CLI tool.
- Engram initializes its structured memory index, loading the primary role, workflows, and rules into the active context.
- During the interaction, the server monitors the conversation and dynamically retrieves supplementary knowledge or examples as needed.
- As the agent performs tasks, Engram tracks the success of tool trajectories and updates its internal dynamic memory.
- The server automatically archives old or irrelevant data and corrects conflicting memories to ensure the AI remains accurate and focused.
Engram MCP Server Setup
To get started with this component of Openclaw Skills, ensure you have the uv package manager installed. You can then add the server to Claude with the following command:
claude mcp add --scope user engram-server -- uvx --from git+https://github.com/DazhuangJammy/Engram engram-server
For manual control and visualization, use the following CLI commands:
# Start the MCP Server
engram-server serve
# Launch the visualization UI
engram-server ui
Engram MCP Server Data Schema & Taxonomy
Engram organizes its cognitive data into a tiered hierarchy to ensure the AI has the most relevant information without overwhelming the context window:
| Component | Category | Behavior |
|---|---|---|
| Identity | Role | Constant; defines persona and tone. |
| Logic | Workflow | Constant; defines step-by-step thinking processes. |
| Constraints | Rules | Constant; defines coding standards and boundaries. |
| Context | Knowledge | Lazy-loaded; retrieved based on semantic triggers. |
| History | Memory | Dynamic; tracks facts and tool outcomes. |
| Archive | Records | Cold storage; used for long-term consistency checks. |
Engram MCP Server Advanced Features
- Zero vector dependency for a lightweight, dependency-free installation.
- Layered loading index that prioritizes roles and rules while lazily loading large knowledge bases.
- Dynamic memory correction mechanisms to prevent the AI from repeating past mistakes.
- Full support for multi-client integration, including Claude Code, Cursor, and Codex.
- Integrated visualization UI for manual auditing and editing of the agent's long-term memory.
SKILL.md
Loading
Related Openclaw Skills
Dockerized Vision Agent Scraper
A containerized web automation tool that utilizes AI vision to navigate complex sites and bypass graphical CAPTCHAs without external services.
Microsoft Graph (Outlook & Calendar)
A comprehensive Microsoft Graph integration for managing Outlook email and calendar events through AI-driven automation.
Lingxi Real-Time Market Data Skill
A professional financial data integration providing real-time stock market metrics across A-shares, HK, and US markets for AI agents.
MiniMax Token Plan Tool
A lightweight JavaScript implementation for MiniMax API integration providing web search, vision-language analysis, and quota management.
ES Search Skill
An AI-driven search utility that translates natural language queries into structured Elasticsearch API calls for comprehensive news and social media monitoring.
PolyOx NBA API
A specialized NBA data and AI analysis interface featuring Polymarket integration and seamless x402 protocol payments.
Featured*
