What is Mengram?
Mengram is a persistent memory backend for AI agents that stores three types of memory—semantic (facts and preferences), episodic (events and decisions), and procedural (workflows that evolve from failures). It is designed for developers and teams who want their AI agents to retain context, learn from past interactions, and automatically improve workflows across sessions.
How to use Mengram?
Install via pip install mengram-ai or npm install mengram-ai, then run mengram setup to create an account and install Claude Code hooks, or manually set the MENGRAM_API_KEY environment variable. Use the Python/JavaScript client (e.g., Mengram(api_key="om-...")) or the CLI to add, search, and manage memories.
Key features of Mengram
- Stores three memory types: semantic, episodic, and procedural.
- Procedures automatically evolve when failures are detected.
- Cognitive profile generates a system prompt from all memories.
- Multi-user isolation with a single API key.
- Imports existing data from ChatGPT, Obsidian, or markdown files.
- Integrates with Claude Code, MCP, LangChain, CrewAI, and OpenClaw.
Use cases of Mengram
- Give Claude Code zero-config memory across sessions with auto-save and recall.
- Build a DevOps agent that learns from deployment failures and improves workflows.
- Create a customer support agent that remembers returning users and past issues.
- Develop a personal assistant that maintains a cognitive profile and chat history.
评论