a community oriented 1:1, vLLM-alike (Continuous batching, paged KV) engine in C++ with additional features (GGUF, RadixAttention, Cache-aware scheduling, ...)
LLM GitHub Repositories (2026) · Open Source Tools
Discover the best LLM open source repositories on GitHub. Explore trending tools, live star counts, source code, and developer projects updated daily.
20+ high-performance LLMs with recipes to pretrain, finetune and deploy at scale.
☕ Community-driven Java implementation of TOON
Give your AI agent a real browser — with a human in the loop. Open-source MCP-native browser agent.
Run Laya, the open-source Jev-compatible System-1 decision model, from Node.js / TypeScript via ONNX Runtime
The complete post-training stack for LLMs, plus RL for any problem. SFT, DPO and multi-turn agentic RL, with evolutionary hyperparameter optimization built in.
An open-source AI agent hub for your own machine: build organizations of collaborating agents that message each other, run in the background around the clock, and share your existing AI subscriptions.
Physical AI runtime for embodied agents — governed action, verified experience, physical memory, and skill evolution.
A living AI-agent office on your desktop wallpaper — Claude Code agents that walk, work, delegate, learn & hold meetings. Per-agent swappable models (Claude/GLM/DeepSeek/Qwen/Kimi/OpenAI/Gemini/Groq/Ollama…), workflows, plugins, voice & Telegram/Discord/LINE. Open source.
Java Inference Tornado toolkit: High performance LLM inference & serving for the JVM on GPUs
RTP-LLM: Alibaba's high-performance LLM inference engine for diverse applications.
:metal: TT-NN operator library, and TT-Metalium low level kernel programming model.
Practical productivity tools for Claude Code, Codex-CLI, and similar CLI coding agents.
Evidence validation and policy engine for AI-assisted biological curation
Fast Rust launcher for DeepSeek Harness, with clear Claude Code/Codex migration, model setup and lightweight local tools
mu (μ): most of what a coding agent decides isn't code. mu hands 40 kinds of yes/no calls to a small model: fewer wasted tokens, fewer approvals on autopilot, fewer runs that stop halfway. Built on pi.
Cost tracking for DeepSeek Harness: session and model costs, token usage, budgets, provider balances, and coding-plan quotas. Bilingual English/Chinese UI.
MCP server for KiCad, connecting AI agents to schematic and PCB workflows, ERC/DRC validation, DFM checks, BOM analysis, and manufacturing preparation—with controlled edits and human-gated release workflows.
Make your computer a home for your AI partner. Avibe is the local-first Agent OS: Vibey, its built-in AI partner, works out of the box, and Claude Code, Codex & OpenCode run unchanged beside it — reach them from your browser, phone, or any chat app.
My Pi coding agent setup: secret redaction, LSP, searchable context, recall, telemetry, observability, and team mode. Toggle with /extensions.