kayacancode/infinite-context-mcp

このページは AI により自動翻訳されています。技術的な詳細は原文をご確認ください。原文を見る

11 stars 4 forks 品質 64 トレンド 8

Unified Context MCP is a memory and retrieval layer for AI agents that turns conversations and tool outputs into structured, reusable context.

概要

A Memory System for AI assistants that goes beyond traditional RAG. Store conversations, build knowledge graphs, extract facts, and maintain user profiles—all searchable with semantic understanding.Use across your apps and LLMs. - - Automatically compress long conversations while preserving important information - - Semantic search across all saved conversations using Pinecone - - Store up to ~5,000 words per chunk with formatting preserved - - LLM-powered intelligent summarization - - Automatically learns your interests, projects, and preferences - - Build knowledge graphs connecting people, projects, concepts - - Extract and chain atomic facts from conversations - - Understand recency and relevance over time - - Combine semantic similarity with temporal and entity signals - - Understand intent (search, save, question, etc.

README

以下は原文 README の抜粋です。原文を見る。

Unified Context Layer MCP

Features

Core Memory

Infinite Context - Automatically compress long conversations while preserving important information
Vector Search - Semantic search across all saved conversations using Pinecone
Full Content Storage - Store up to ~5,000 words per chunk with formatting preserved
Smart Compression - LLM-powered intelligent summarization

Memory System

User Profiles - Automatically learns your interests, projects, and preferences
Entity Graphs - Build knowledge graphs connecting people, projects, concepts
Fact Extraction - Extract and chain atomic facts from conversations
Temporal Awareness - Understand recency and relevance over time
Hybrid Scoring - Combine semantic similarity with temporal and entity signals

Query Understanding

Query Classification - Understand intent (search, save, question, etc.)
Query Rewrites - Generate synonyms, broader terms, and expansions
Enhanced Search - Guardrails, auto-refinement, and follow-up recommendations

Indexing

GitHub Repositories - Index entire repos for code search
Documentation Sites - Crawl and index docs
Websites - Full website crawling
Local Filesystems - Index local directories
Single URLs - Index individual pages (blogs, ChatGPT conversations, tweets)

🚀 Quick Start

1. Clone the Repository

git clone https://github.com/yourusername/infinite-context-mcp.git
cd infinite-context-mcp

2. Create Virtual Environment

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install Dependencies

pip install -r requirements.txt

4. Configure API Keys

cp env_example.txt .env

Edit .env with your API keys:

ANTHROPIC_API_KEY=your_anthropic_api_key_here
OPENAI_API_KEY=your_openai_api_key_here
PINECONE_API_KEY=your_pinecone_api_key_here
PINECONE_INDEX_NAME=infinite-context-index  # Optional, this is the default

Get your API keys:

Anthropic Console - For Claude (AI responses)
OpenAI Platform - For embeddings
Pinecone Console - For vector storage (free tier available)

5. Update the Run Script

Edit run_mcp.sh to point to your installation:

#!/bin/bash
cd /path/to/your/infinite-context-mcp
source venv/bin/activate
exec python main.py

Make it executable:

chmod +x run_mcp.sh

6. Connect to Your AI Tool

For Cursor

Add to your Cursor MCP settings (~/.cursor/mcp.json or via Settings > MCP):

{
  "mcpServers": {
    "infinite-context": {

GitHub で README 全文を見る

インストール

This server does not publish a one-line install command.

Open the repository installation guide MCP 初心者？mcpskill.org のガイドから始めましょう →

設定

{
  "mcpServers": {
    "infinite-context": {
      "command": "/path/to/your/infinite-context-mcp/run_mcp.sh",
      "args": [],
      "env": {
        "PYTHONPATH": "/path/to/your/infinite-context-mcp"
      }
    }
  }
}