EGC logo

EGC

Free

跨会话持久记忆 MCP 服务,支持 Claude Code、Cursor、Gemini CLI、Codex、Windsurf 等多种 AI 编程工具,SQLite 本地存储,上下文重置后状态保持。提供 egc-memory(14 工具)与 egc-guardian(5 工具)。`npm install -g @egchq/egc`

FreeFree tier
Type
Open Source

About EGC

EGC (Every Agent Gets Context) is an open-source persistent memory and intelligence layer for AI coding agents. It provides a shared brain that automatically synchronizes memory, skills, and live context across 20+ AI coding tools including Cursor, Claude Code, GitHub Copilot, Aider, Codex, Windsurf, Gemini CLI, and more. With zero configuration, EGC eliminates the need to rebuild context or repeat instructions across sessions, tabs, or terminals. It supports multiple LLMs natively (Claude, GPT-4o, Gemini, DeepSeek, Mistral, Groq, Cohere, Vertex AI) and via OpenRouter for models like Qwen3 and Llama 4. EGC reduces context waste by up to 90% and cuts token costs. It includes a built-in Guardian safety layer that validates commands before execution, blocks dangerous writes, and detects prompt injection. Memory remains local and encrypted, never committed to git. Installation is a single npm command: npm install -g @egchq/egc.

Key Features

Persistent project memory shared across all AI agents and sessions
Supports 20+ AI coding tools: Cursor, Claude Code, Copilot, Aider, Codex, Windsurf, Gemini CLI, and terminal agents
Native support for multiple LLMs: Claude, GPT-4o, Gemini, DeepSeek, Mistral, Groq, Cohere, Vertex AI, plus OpenRouter
Guardian safety layer: command validation, dangerous write blocking, prompt injection detection
Natural language interface – speak naturally to save, recall, or manage context
Local, encrypted storage that never leaves your machine
Zero configuration setup with a single npm install command
Reduces context waste by up to 90% and lowers token costs

Pros & Cons

Pros
  • Eliminates need to repeat project context across different AI tools and sessions
  • Reduces token usage and API costs significantly (up to 90% less context waste)
  • Works out of the box with 20+ popular AI coding tools without configuration
  • Local, encrypted storage ensures privacy and security
  • Guardian feature provides an additional safety layer against dangerous commands and prompt injection
  • Supports a wide variety of LLMs, including open-source models via OpenRouter

Best For

Maintaining project context across multiple AI coding tools in the same projectReducing token consumption and API costs by avoiding context repetitionEnforcing safety guardrails for AI-generated commands in development environmentsBuilding a persistent shared memory for team collaboration with AI agentsAutomatically synchronizing decisions and learnings between IDE agents and terminal agents