ultracontext/ultracontext

Open Source Context infrastructure for AI agents. Auto-capture and share your agents' context everywhere.

What it solves

UltraContext provides a shared, real-time context layer for AI agents. It solves the problem of context fragmentation, where information known by one agent (like Claude Code or Codex) is unavailable to others, preventing users from seamlessly continuing sessions or sharing knowledge across different AI tools.

How it works

The system captures agent context in real-time and makes it available to all other agents via a centralized infrastructure. It uses a Model Context Protocol (MCP) server to give agents awareness of other agents' states. For developers, it provides a git-like Context API that allows for storing, versioning, and retrieving context with primitives like create, append, update, and delete.

Who it’s for

  • AI Agent Users: People using multiple AI coding agents or tools and wanting to move between them without losing progress.
  • AI Developers: Builders who want a framework-agnostic way to manage, version, and "time-travel" through agent context history.

Highlights

  • Real-time Sync: Automatically ingests sessions from tools like Claude Code and Codex.
  • ** uma MCP Server**: Enables agents to share context across different platforms.
  • Git-like Versioning: Every change to the context creates a new version, allowing for history tracking and time-travel.
  • Framework Agnostic: Works with any LLM framework via JavaScript/TypeScript and Python SDKs.

Related

  • Project
    rodrigo-arenas/Sklearn-genetic-opt

    A scikit-learn-compatible optimization toolkit that uses genetic algorithms for hyperparameter tuning and feature selection, offering a more efficient alternative to grid and random search.

  • Project
    cisco-open/network-sketcher

    An AI-native network design tool that generates L1/L2/L3 topology diagrams and device tables from structured data or natural language workflows.

  • Project
    charlie12345/ROCmFPX

    An extension for llama.cpp that introduces AMD-optimized GGUF quantization formats and accelerated kernels to improve LLM inference speed and memory efficiency on Radeon GPUs.

  • Project
    cloudwego/abcoder

    An AI-oriented code-processing framework that uses a Universal Abstract-Syntax-Tree (UniAST) to provide LLMs with precise, structured coding context for better AI-assisted programming.