* [Langfuse v4.21.0](https://github.com/langfuse/langfuse) – LLM engineering platform for collaboratively developing, monitoring, evaluating, and debugging AI applications.

* [Ollama v0.33.0](https://github.com/ollama/ollama) – Tool for running and managing large language models.

* [Dify 1.17.0](https://github.com/langgenius/dify) – Platform for developing large language model applications with AI workflows, model management, and observability.

* [CopilotKit v1.69.0](https://github.com/CopilotKit/CopilotKit) – Tool for building integrated AI assistants and agents that operate within user applications.

* [Agent Canvas v1.15.0](https://github.com/OpenHands/OpenHands) – Self-hosted developer control center that runs coding agents across local, remote, or cloud backends and automates everyday engineering workflows.

* [LocalAI v4.9.0](https://github.com/mudler/LocalAI) – Self-hosted alternative to popular AI APIs for local inferencing on consumer-grade hardware.

* [FinanceMCP (Synapse) - MCP v4.11.0](https://github.com/guangxiangdebizi/FinanceMCP) – MCP integrating Tushare and Binance APIs to provide real-time multi-asset financial data and news for language models.

* [opencodex v2.33.0](https://github.com/lidge-jun/opencodex) – Lightweight local proxy that translates Codex Responses API into multiple LLM providers with full support for streaming, tools, reasoning tokens, and images.

* [TencentDB Agent Memory v2.0.1](https://github.com/TencentCloud/TencentDB-Agent-Memory) – TencentDB Agent Memory provides symbolic short-term and layered long-term memory for AI agents to retain context and reduce token usage without external APIs.

* [Aiden v4.21.0](https://github.com/taracodlabs/aiden) – Local-first AI operating system for Windows, Linux, WSL, and macOS with 1,400+ skills, 80+ tools, and provider routing.

* [avoid-ai-writing v3.26.0](https://github.com/conorbronsdon/avoid-ai-writing) – Audit and rewrite content to remove AI writing patterns, with detect-only or edit-in-place modes and configurable voice profiles for agent skills.

* [LLM Gateway v1.14.0](https://github.com/theopenco/llmgateway) – API gateway that routes and manages requests across multiple LLM providers with an OpenAI-compatible API, usage analytics, and cost tracking.

* [FastGPT v4.16.1](https://github.com/labring/FastGPT) – Knowledge-based platform leveraging LLMs for data processing, retrieval, and AI workflow orchestration.

* [LangChain langchain-fireworks=...](https://github.com/langchain-ai/langchain) – Framework for building agents and LLM-powered applications with modular components and integrations.

* [Reasonix studio-v2.9.0](https://github.com/esengine/DeepSeek-Reasonix) – DeepSeek-native AI coding agent for your terminal, engineered around prefix-cache stability to keep token costs low across long sessions.

* [SQLBot v1.10.1](https://github.com/dataease/SQLBot) – RAG-based conversational data analysis and Text-to-SQL generation using large language models.

* [Promptfoo 0.122.1](https://github.com/promptfoo/promptfoo) – Developer-friendly local tool for testing, evaluating, and securing large language model applications.

* [Kortix v0.13.6](https://github.com/kortix-ai/suna) – Autonomous company OS where Linux-sandboxed AI agents run code, manage files, and automate operations 24/7.

* [Bifrost plugins/telemetry/v1...](https://github.com/maximhq/bifrost) – High-performance AI gateway offering multi-provider access with automatic failover, load balancing, and zero-downtime deployments.

* [Headroom v0.9.0](https://github.com/gglucass/headroom-desktop) – Local-first desktop tray app that routes coding clients through a token-saving optimization pipeline and shows savings analytics, primarily for Claude Code.

* [Abu v0.42.0](https://github.com/PM-Shawn/Abu-Cowork) – Privacy-first desktop AI coworking app with multi-model support, agent skills, IM integration, MCP/Playwright, scheduling, and built-in security controls.

* [NoteGen note-gen-v0.36.0](https://github.com/codexu/note-gen) – Cross-platform Markdown note-taking application using AI to organize fragmented knowledge into readable notes.

* [Mastra @mastra/core@1.62.0](https://github.com/mastra-ai/mastra) – Opinionated TypeScript framework for building AI applications and features quickly.

* [freebuff-proxy v1.2.0](https://github.com/trefeon/freebuff-proxy) – High-performance OpenAI-compatible gateway and protocol bridge that manages sessions, pools tokens, and forwards chat completions with SSE streaming and TLS stealth.

* [Prompt Optimizer v2.11.9](https://github.com/linshenkx/prompt-optimizer) – Tool for enhancing prompt quality to improve AI interaction results.

* [Gerbil v1.28.0](https://github.com/lone-cloud/gerbil) – Desktop app for running Large Language Models locally with cross-platform support and integrated image generation.

* [Sokuji v0.39.0](https://github.com/kizuna-ai-lab/sokuji) – Desktop application providing live speech translation and audio routing using multiple AI-powered APIs.

* [yzma v1.25.0](https://github.com/hybridgroup/yzma) – Go-based library for hardware-accelerated local inference with llama.cpp integration.

* [Ax 24.0.8](https://github.com/ax-llm/ax) – End-to-end streaming framework for building multi-modal agents with typed signatures.

* [Read Frog v1.46.5](https://github.com/mengxi-ream/read-frog) – AI-powered browser extension for immersive language learning with translation and article analysis.

* [easy-stock v0.9.0](https://github.com/jundizhou/easy-stock) – Local-first A 股行情分析与AI智能投研代理，支持日常共识与短线研判。

* [Vellium v1.1.0](https://github.com/tg-prplx/vellium) – Desktop chat app using Electron, local Express API, and SQLite for roleplay and creative-writing LLM workflows.

* [Memmy v1.1.0](https://github.com/MemTensor/memmy-agent) – Memory hub and local agent runtime that provides shared long-term personal context across multiple AI agents.

* [llm.rb v15.1.0](https://github.com/r-uby-dev/llm) – Advanced Ruby AI runtime for building capable applications with multiple LLM providers, streaming, tool calls, and RAG support.

* [Cherry Studio v2.0.9](https://github.com/CherryHQ/cherry-studio) – Desktop client supporting multiple LLM providers, available on Windows, Mac, and Linux.

* [yibiao-simple v2.25.19](https://github.com/FB208/OpenBidKit_Yibiao) – Out-of-the-box AI bid writing and checking tool with knowledge base and duplicate detection.

* [MDFlux v0.3.0](https://github.com/ibrahimqureshae/mdflux) – MDFlux is a local-first desktop tool that converts many document types and scanned PDFs into clean, AI-ready Markdown with OCR and optional cleanup.

* [CortexDB v2.74.0](https://github.com/liliang-cn/cortexdb) – Pure-Go, single-file embedded AI memory and knowledge graph library that provides hybrid vector and lexical retrieval, RDF/SPARQL graph features, and MCP tools via SQLite.

* [llama\_cpp.rb v0.28.0](https://github.com/yoshoku/llama_cpp.rb) – Ruby bindings for llama.cpp, enabling easy integration of the library in Ruby applications.

* [AI SDK @ai-sdk/deepgram@3.1...](https://github.com/vercel/ai) – TypeScript toolkit for building AI-powered applications with popular frameworks.

* [obsidian-mcp-server v3.5.0](https://github.com/cyanheads/obsidian-mcp-server) – MCP server for Obsidian vaults that reads, writes, searches, and surgically edits notes, tags, and frontmatter via the Local REST API plugin.

* [Ix v0.10.0](https://github.com/ix-infrastructure/Ix) – Command-line tool that parses a codebase into a persistent queryable symbol, call, and import graph for efficient AI and human reasoning.

* [Open Multi-Agent v1.16.1](https://github.com/open-multi-agent/open-multi-agent) – TypeScript-native multi-agent orchestration that decomposes goals into parallel task DAGs and synthesizes results.

* [NobodyWho nobodywho-flutter-v3...](https://github.com/nobodywho-ooo/nobodywho) – Inference engine for running LLMs locally and efficiently on any device.

* [OpenOPC-Shadow-Adapter v1.0.0](https://github.com/AhmadHassan-BTed/OpenOPC-Shadow-Adapter) – High-concurrency Temporal bridge for OpenOPC that enables non-blocking human sign-off and remote BYOC execution via a shadow portal and SQLite WAL state parking.

* [overtchat v0.17.0](https://github.com/yoloyash/overtchat) – Lightweight self-hosted chat client focused on fast startup, low resource use, and BYO API keys or local models, with optional web search and mobile support.

* [OpenScience v2.0.49](https://github.com/synthetic-sciences/openscience) – Model-agnostic AI workbench that loops through literature review, hypothesis building, code and experiment execution, and research write-ups in a browser workspace.

* [MCP Mesh v4.276.0](https://github.com/decocms/studio) – Control plane routing MCP client traffic through one governed endpoint with auth, RBAC, policy enforcement, and observability.

* [RunAnywhere v0.20.29](https://github.com/RunanywhereAI/runanywhere-sdks) – On-device mobile SDKs for running LLMs, speech-to-text, and text-to-speech locally.

* [Token Optimizer MCP v5.7.1](https://github.com/ooples/token-optimizer-mcp) – Model Context Protocol server that optimizes Claude Code context usage using caching, Brotli compression, and smart tool replacements for large text and file operations.

* [PasteGuard v0.9.3](https://github.com/sgasser/pasteguard) – OpenAI-compatible privacy proxy that masks personal data and secrets before they reach external LLMs.

* [AionUi v2.1.61](https://github.com/iOfficeAI/AionUi) – Modern GUI for multi-model AI chat with integrated file management and Excel processing.

* [HelixML 2.12.6](https://github.com/helixml/helix) – Private GenAI stack for deploying AI agents with support for RAG, API calls, vision, and efficient GPU scheduling.

* [Hermes Studio v0.6.47](https://github.com/EKKOLearnAI/hermes-studio) – Hermes Studio is a desktop app, local runtime, and web console for managing Hermes Agent chat, models, profiles, automation jobs, and analytics while keeping everything local.

* [MemOS memos-local-plugin-v...](https://github.com/MemTensor/MemOS) – Memory operating system for LLMs and AI agents providing unified store/retrieve/manage, multi-modal long-term memory and skill reuse.

* [or v0.6.16](https://github.com/ktsoator/or) – Modular Go toolkit with provider-neutral language model access and stateful agent tool loops with typed streaming events.

* [Libre WebUI v0.28.0](https://github.com/libre-webui/libre-webui) – Privacy-first self-hosted chat interface connecting to local and cloud AI providers with a plugin system and no telemetry or tracking.

* [Super-Agent-Party v0.4.3](https://github.com/heshengtao/super-agent-party) – 3D AI desktop companion platform with modular LLM enhancements, multi-terminal deployment, and cross-platform compatibility.

* [liter-llm v1.18.1](https://github.com/xberg-io/liter-llm) – A lighter, faster, safer universal LLM API client built with a Rust core and many native language bindings, with a drop-in OpenAI-compatible proxy.

* [SigMap v8.28.1](https://github.com/manojmallick/sigmap) – Finds relevant files in a codebase and supplies compact function and class signatures to AI models to reduce tokens and improve answers.

* [Pinvou Agent v0.8.6](https://github.com/Pinvou/pinvou-agent) – Desktop AI agent workspace that connects models with tools and personal knowledge to produce and manage real deliverable artifacts.

* [Whisplay-AI-Chatbot v2.1.2](https://github.com/PiSugar/whisplay-ai-chatbot) – Pocket-sized AI chatbot device built for Raspberry Pi Zero 2w with push-to-talk voice interaction, wake-word, and image generation.

* [ggrun v3.2.8](https://github.com/raketenkater/ggrun) – Auto-tuned launcher that measures multi-GPU hardware for GGUF models, picks an optimal llama.cpp/ik\_llama.cpp backend, and serves an OpenAI-compatible API.

* [DEEIX Chat v0.3.6](https://github.com/DEEIX-AI/DEEIX-Chat) – Enterprise AI workspace unifying model routing, multimodal chat, file/RAG workflows, tools, billing, identity, and operations.

* [RepoWiki v0.3.3](https://github.com/he-yufeng/RepoWiki) – RepoWiki generates structured wiki documentation for local or GitHub codebases from a CLI or web interface, exporting Markdown, JSON, or self-contained HTML with terminal Q&A.

* [go-llama v0.2.3](https://github.com/goccy/go-llama) – Pure Go inference engine for GGUF models, built from llama.cpp compiled to WebAssembly and translated to Go without wasm runtime.

* [AgentField v0.1.135](https://github.com/Agent-Field/agentfield) – Control-plane infrastructure that deploys, scales, and secures autonomous AI agents as observable, identity-aware backend services.

* [OpenBiliClaw extension-v0.3.212](https://github.com/whiteguo233/OpenBiliClaw) – Local, private, self-improving content discovery agent that learns from interactions to find videos and posts across Bilibili, Xiaohongshu, Douyin, YouTube, and more.

* [ElevenLabs Monorepo for NPM Package @elevenlabs/react@1....](https://github.com/elevenlabs/packages) – Monorepo managing multiple npm packages under the @elevenlabs scope.

* [QVAC sdk-v0.18.2](https://github.com/tetherto/qvac) – Local-first, cross-platform SDK for building peer-to-peer AI apps with local model inference, speech, translation, and RAG.

* [Noobot v4.2.5](https://github.com/xiayu1987/noobot) – Self-hosted AI agent workspace providing isolated sessions, tool calling, multi-model routing, MCP connectors, and multi-agent workflows via web and desktop clients.

* [react-native-litert-lm v0.6.1](https://github.com/hung-yueh/react-native-litert-lm) – React Native library for high-performance on-device LLM inference with LiteRT-LM and Nitro Modules, featuring crash-free memory handling and structured outputs.

* [WhatIfIBought v2.0.4](https://github.com/mamawai/wtfibought) – Live arena where LLM trading agents trade real market data with virtual money while exposing every thought, tool call, and post-trade review.

* [Wegent wework-v0.2.6](https://github.com/wecode-ai/Wegent) – Self-hostable platform for building and running AI agent teams across chat, coding, knowledge, and automation with shared capabilities and local execution via Wework.

* [San v1.22.5](https://github.com/genai-io/san) – Terminal-native unified runtime for specialized AI agents built on pluggable LLMs, search backends, personas, and skills.

* [Ferro Labs AI Gateway v1.4.5](https://github.com/ferro-labs/ai-gateway) – High-performance Go-based AI gateway routing OpenAI-compatible LLM requests across 30+ providers with caching, guardrails, and cost controls.

* [GoAI v0.9.7](https://github.com/zendev-sh/goai) – Go SDK providing a unified API for 25+ AI providers with streaming, structured output, and MCP support.

* [Lattice v1.14.8](https://github.com/Protocol-Lattice/go-agent) – Go agent framework with graph-aware memory, UTCP-native tools, and multi-agent orchestration.

* [dsh-commandcode-provider v0.6.2](https://github.com/Mars-Sea/dsh-commandcode-provider) – Unofficial DeepSeek Harness provider plugin that adds a Command Code provider route with live model catalog, plan/deal annotations, and reasoning-effort plus vision support.

* [Context7 MCP @upstash/context7-mc...](https://github.com/upstash/context7) – Fetches up-to-date documentation and code examples for prompts in LLMs.

* [Marginalia v0.3.6](https://github.com/shenmintao/marginalia) – Local-first research agent that ingests messy personal knowledge, keeps it in a folder tree, reads original source windows, and generates cited answers with durable notes.

* [Blade Code v0.10.98](https://github.com/echoVic/blade-code) – AI CLI and web UI coding agent with 20+ tools, MCP support, and configurable providers for model-driven coding workflows.

* [Coddy 0.9.79](https://github.com/coddy-project/coddy-agent) – Coddy is a distroless-friendly Go-based general-purpose agent harness with a ReAct loop, ACP/HTTP/Telegram interfaces, MCP and skills integration, and long-term memory.

* [llama.rn v0.13.0-rc.1](https://github.com/mybigday/llama.rn) – React Native binding for running LLaMA model inference with multimodal support including vision and audio.

* [Off Grid AI v0.0.43](https://github.com/off-grid-ai/OGAD) – Local-first AI runtime with a studio and an OpenAI-compatible gateway, running open multimodal models entirely on-device with an optional always-on assistant layer.

* [nano banana pro🍌 v0.9.0-rc.4](https://github.com/Anionex/banana-slides) – AI-native PPT generation app that turns ideas, outlines, documents, and images into editable PPTX with template control and conversational refinement.

* [pgEdge Postgres MCP Server and Natural Language Agent v1.1.0-beta3](https://github.com/pgEdge/pgedge-postgres-mcp) – PostgreSQL MCP server enabling SQL queries from MCP-compatible clients, with a natural-language agent plus CLI and web UIs.

* [senpi v2026.8.23](https://github.com/code-yeongyu/senpi) – Opinionated TypeScript monorepo fork of pi-mono that provides a curated coding-agent runtime with builtin extensions and core tweaks.

* [warpdrv v0.6.17-beta](https://github.com/mikjee/warpdrv) – Desktop toolkit for managing llama-server instances and chatting with local LLMs, featuring a rich UI, voice, MCP tools, and workflow helpers like RAG and agents.

* [AgentSphere v1.0.70-alpha](https://github.com/nullpointexception-i/agent-sphere) – LLM-driven AI agent orchestration platform that coordinates a perception–planning–execution–feedback loop with tool, MCP, CLI, and browser capabilities.

* [DIAL Chat 1.0.0-rc.6](https://github.com/epam/ai-dial-chat) – Default UI for DIAL offering conversation interface, IDP support, model side-by-side comparison, extensions and theming.

* [Tingly Box v0.260827.0-rc1](https://github.com/tingly-dev/tingly-box) – High-performance desktop LLM proxy that unifies access to many models and providers via a single OpenAI-compatible API.

* [NodeTool v0.7.1-nightly.20260...](https://github.com/nodetool-ai/nodetool) – Visual platform for building and deploying AI workflows, agents, and multimodal pipelines via drag-and-drop nodes.

* [Stately Agent @statelyai/agent@2.0...](https://github.com/statelyai/agent) – State-machine-powered LLM agent logic built on XState, where model decisions and tool effects run within explicit, inspectable control flow.