关于
Contextra is an MCP (Model Context Protocol) Memory Server designed to equip AI agents with persistent memory. It allows agents to learn and recall past interactions and information, functioning as a structured storage and retrieval system. The system employs a two-stage retrieval workflow, optimizing token usage by first providing memory summaries for agent preview, followed by selective retrieval of full content, ensuring efficient access to comprehensive information.
主要功能
- Employs an agent-driven architecture, keeping cognitive processing with the AI agent.
- Includes MCP-native guidance through built-in prompts for effective agent usage.
- Offers five canonical MCP tools for memory operations (write, search, retrieve, list, inspect).
- Provides local-first storage for full data control and privacy.
- 0 GitHub stars
- Features a lightweight-first design utilizing SQLite and ChromaDB, with no external dependencies.
使用案例
- Enhancing GitHub Copilot's memory and contextual understanding within VS Code.
- Optimizing token usage for AI agents by implementing a two-stage memory retrieval process.
- Enabling AI agents to store, search, and retrieve relevant information from past interactions.