100 Must-Have GitHub Repos
for 2025–2026

After digging through ByteByteGo roundups, GitHub trending pages, and a dozen "awesome-AI" lists, here's the curated set of 100 repositories that actually matter in 2025–2026 — not demo-ware, not abandoned side projects, but tools people are shipping production work with.

🤖 1. AI Agents & Automation

The category that exploded fastest this cycle. Visual builders, autonomous coding agents, and crews of specialized bots now make up half of the top trending repos on GitHub.

• OpenClaw★ 210k+
  The breakout star of 2026, created by PSPDFKit founder Peter Steinberger — surged from 9,000 to over 60,000 stars in days. github.com/OpenClaw/OpenClaw
• awesome-ai-agents-2026
  177 copy-paste SOUL.md configs for OpenClaw, covering workflow automation, web scraping, and proactive scheduling. github.com/caramaschiHG/awesome-ai-agents-2026
• n8n
  Open-source workflow automation with a visual no-code interface, native AI nodes, and 400+ integrations. Self-hosted, fair-code licensed. github.com/n8n-io/n8n
• CrewAI
  Treats agents as crew members with defined roles and goals. Production-ready since late 2025 with enterprise observability and scheduling, model-agnostic across GPT, Claude, Gemini and local Llama via Ollama. github.com/crewAIInc/crewAI
• AutoGPT
  The original autonomous agent — chains LLM calls to complete long-horizon tasks without hand-holding. Still massively starred. github.com/Significant-Gravitas/AutoGPT
• Browser Use★ 97.7k
  Agent and CLI for reliable website automation — form filling, data retrieval, custom tools. github.com/browser-use/browser-use
• Daytona★ 72.5k
  Secure, elastic sandbox infrastructure for AI-generated code, controllable via SDK, API, and CLI. github.com/daytonaio/daytona
• DeerFlow (ByteDance)
  Ground-up rewrite for long-horizon agents — sub-agents, memory, sandboxes, skills, context engineering. Hit #1 on GitHub Trending in Feb 2026. github.com/bytedance/deer-flow
• MetaGPT
  Multi-agent framework where agents take on roles (CEO, engineer, PM) to collaborate on software via SOPs. github.com/geekan/MetaGPT
• AutoGen (Microsoft)
  Framework for conversational multi-agent workflows where agents talk to each other to solve complex tasks. github.com/microsoft/autogen
• Open Interpreter
  Gives an LLM access to your terminal to write and execute code locally — like a Code Interpreter with no upload limits, supporting OpenAI, Anthropic, and Ollama models. github.com/OpenInterpreter/open-interpreter
• SWE-agent
  Princeton's agent that autonomously resolves GitHub issues and bugs — one of the best open coding agents around. github.com/princeton-nlp/SWE-agent
• Mastra
  TypeScript-native agent framework, the go-to for teams building AI products in JS/TS with type safety. github.com/mastra-ai/mastra
• Langflow★ 146k
  Visual builder for prototyping and deploying agents without writing extensive code. github.com/logspace-ai/langflow
• Flowise
  Drag-and-drop UI for LangChain flows — easy to self-host, great for RAG pipelines. github.com/FlowiseAI/Flowise
• Dify★ 136k
  Visual workflow builder for AI automations wired into existing tools, with a strong self-hosted story. github.com/langgenius/dify

🖥️ 2. Local LLMs & Ollama Ecosystem

Local inference matured into a full ecosystem — runtimes, serving engines, and polished desktop clients that rival cloud chat products.

• Ollama★ 165k+
  Did for local LLMs what Docker did for containers — one command to pull and run Llama, DeepSeek, Mistral, Gemma and more. github.com/ollama/ollama
• Open WebUI
  Ollama's perfect companion — the backbone of the local AI movement, a fully self-hosted chat UI. github.com/open-webui/open-webui
• llama.cpp
  Run LLaMA and other LLMs on CPU/GPU with blazing efficiency — the engine under many local AI tools. github.com/ggerganov/llama.cpp
• vLLM
  High-throughput serving engine with PagedAttention for production-grade local inference. github.com/vllm-project/vllm
• DeepSeek-V3
  Proof open models can compete with proprietary ones — free for commercial use, fine-tunable, runs locally via Ollama. github.com/deepseek-ai/DeepSeek-V3
• LocalAI
  Drop-in OpenAI API replacement for self-hosted models — run any GGUF model behind an OpenAI-compatible REST API. github.com/mudler/LocalAI
• LM Studio
  Desktop app + API for running local models easily — the best GUI experience for GGUF/Ollama models. github.com/lmstudio-ai
• Jan
  Privacy-focused, fully offline ChatGPT alternative that runs entirely on-device. github.com/janhq/jan
• GPT4All
  Run LLMs 100% locally with a simple desktop app — no GPU required. github.com/nomic-ai/gpt4all
• Yao Agents
  Local-first execution platform with Docker sandbox isolation, BYOK config, MCP support, and multi-platform messaging (WeChat, Feishu, Telegram, Discord). github.com/YaoApp/yao

🎨 3. Image Generation

• ComfyUI
  Node-based interface giving advanced, fully customizable control over diffusion image-generation workflows. github.com/comfyanonymous/ComfyUI
• AUTOMATIC1111 (Stable Diffusion WebUI)
  The simplicity pick — install, load a model, start generating in minutes. github.com/AUTOMATIC1111/stable-diffusion-webui
• InvokeAI
  Professional creative tool for Stable Diffusion with a polished canvas UI and model management. github.com/invoke-ai/InvokeAI
• Fooocus
  The simplest possible SDXL interface — focus on prompting, not configuration. github.com/lllyasviel/Fooocus
• FLUX.1 (Black Forest Labs)
  The hottest open image model of 2025–2026 — state-of-the-art quality, fully open weights. github.com/black-forest-labs/flux
• Real-ESRGAN
  AI-powered image upscaling and restoration — sharpens any low-res image. github.com/xinntao/Real-ESRGAN
• Open-Generative-AI
  Open alternative to commercial AI image/video platforms with 200+ models (Flux, Kling-style, Veo-style), MIT licensed. github.com/anil-matcha/open-generative-ai
• Stable Diffusion XL Turbo
  Near real-time generation from SDXL — 1–4 steps, near-instant results.

🎬 4. Video Generation

• Open-Sora 2.0
  Open-source video generation now competing with commercial alternatives like Runway Gen-4.5 and Kling 3.0. github.com/hpcaitech/Open-Sora
• HunyuanVideo (Tencent)★ 11.9k
  One of the stronger open video generators, handling 720p and 1080p. github.com/Tencent/HunyuanVideo
• CogVideoX-1.5
  Apache 2.0, loads natively via Hugging Face Diffusers — smooth transitions, strong prompt following. github.com/THUDM/CogVideo
• Wan2.1★ 13.2k
  Open, large-scale video generative models. github.com/Wan-Video/Wan2.1
• SadTalker / LivePortrait★ 18.4k
  Brings portraits to life with face animation and video generation. github.com/OpenTalker/SadTalker
• Digital Human / AI Avatar Toolkit★ 12.7k
  Open-source digital human toolkit for offline avatar generation and cloning. github.com/digital-avatar
• Deforum Stable Diffusion
  The top choice for AI music videos and prompt-driven animation. github.com/deforum-art/deforum-stable-diffusion
• Text-To-Video-AI
  Topic in, script + images + narration out, using OpenAI and Pexels APIs across multiple languages. github.com/SamurAIGPT/Text-To-Video-AI

🗣️ 5. Text-to-Speech & Voice

• F5-TTS
  Open TTS that rivals commercial APIs on zero-shot voice cloning, built on flow matching with a DiT transformer — clones a voice from 5–15 seconds of audio. github.com/SWivid/F5-TTS
• Fish Speech (S2)
  SOTA voice cloning with cross-lingual generalization across 80+ languages without retraining. github.com/fishaudio/fish-speech
• Coqui TTS
  Robust, well-documented open TTS toolkit supporting dozens of voices and languages. github.com/coqui-ai/TTS
• Tortoise TTS
  High-quality, expressive multi-voice TTS — slower, but stunning for audiobooks and narration. github.com/neonbjb/tortoise-tts
• Suno Bark
  Expressive speech and sound-effects model — one of the truly free open audio generation options. github.com/suno-ai/bark
• WhisperX
  Faster-than-real-time speech-to-text with speaker diarization — the best open STT pipeline. github.com/m-bain/whisperX
• Faster-Whisper
  CTranslate2-optimized Whisper, 4x faster transcription on CPU and GPU. github.com/SYSTRAN/faster-whisper
• ComfyUI-TTS-Audio-Suite
  Engine-agnostic node interface — v4.22.7 added Echo-TTS for premium English cloning plus Qwen3-TTS and CosyVoice3 for multilingual synthesis. github.com/diodiogod/ComfyUI-TTS-Audio-Suite

⚡ 6. Productivity & Personal Tools

• OpenCut
  Free, open-source web video editor — the open alternative to CapCut. github.com/OpenCut-app/OpenCut
• AppFlowy
  Open-source Notion alternative, fully self-hosted, Markdown + database + AI-ready. github.com/AppFlowy-IO/AppFlowy
• Obsidian plugin ecosystem
  The most powerful personal knowledge management ecosystem, extended by hundreds of plugins. github.com/obsidianmd
• Actual Budget
  Free, privacy-focused, self-hosted personal finance and budgeting app. github.com/actualbudget/actual
• Cal.com
  Open-source Calendly alternative, fully self-hosted. github.com/calcom/cal.com
• Docmost
  Open-source Notion + Confluence alternative for teams — wiki, docs, real-time collaboration. github.com/docmost/docmost
• Plane
  Open-source Jira/Linear alternative for project management. github.com/makeplane/plane
• Hoppscotch
  Open-source Postman alternative, self-hosted, test APIs in-browser. github.com/hoppscotch/hoppscotch
• Excalidraw
  Virtual whiteboard for sketching diagrams, system designs, and flowcharts. github.com/excalidraw/excalidraw
• Memos
  Lightweight, self-hosted Twitter-like note-taking app. github.com/usememos/memos
• Stirling-PDF
  Locally hosted PDF manipulation — convert, merge, split, OCR, all offline. github.com/Stirling-Tools/Stirling-PDF
• Screenpipe
  Open-source personal AI memory — records screen and audio 24/7, queryable with AI. github.com/mediar-ai/screenpipe

🔍 7. RAG, Search & Knowledge

• AnythingLLM
  Drop in documents, websites, or databases and chat with them using any LLM, locally or via API. github.com/Mintplex-Labs/anything-llm
• Perplexica
  Open-source Perplexity clone — an AI search engine that actually searches the web and cites sources. github.com/ItzCrazyKns/Perplexica
• Mem0
  The go-to memory layer for production agents, pairing well with LangChain or CrewAI. github.com/mem0ai/mem0
• Chroma
  Open-source AI-native vector database — the default choice for local RAG projects. github.com/chroma-core/chroma
• Qdrant
  High-performance, fully self-hostable vector search engine for production RAG. github.com/qdrant/qdrant
• Crawl4AI
  Web crawler optimized for LLMs and RAG — scrapes and returns clean markdown. github.com/unclecode/crawl4ai
• Firecrawl
  Turns any website into clean LLM-ready data, used by thousands of AI apps. github.com/mendableai/firecrawl

🛡️ 8. Coding Agents & Dev Tools

• Gemini CLI
  Google's agentic coding tool — chat with, instruct, and automate from the terminal via a simple npx command. github.com/google-gemini/gemini-cli
• Aider
  AI pair programmer in your terminal, works with GPT-4, Claude, and local models, edits your codebase directly. github.com/paul-gauthier/aider
• Continue
  Open-source Copilot alternative for VS Code and JetBrains — bring any model, local or cloud. github.com/continuedev/continue
• awesome-cursorrules
  Community-curated best-practice rules for the Cursor editor. github.com/PatrickJS/awesome-cursorrules
• Tabby
  Self-hosted AI coding assistant with an OpenAI-compatible API — zero data sent to the cloud. github.com/TabbyML/tabby
• Codeium
  Free AI code completion across 70+ languages with a generous free tier. github.com/Exafunction/codeium
• Claude Code (Anthropic)
  Excellent at debugging, explaining errors, and step-by-step reasoning — now open-source. github.com/anthropics/claude-code

🌐 9. Scraping, Web & Data

• Playwright
  Microsoft's browser automation library — the gold standard for scraping and E2E testing. github.com/microsoft/playwright
• Puppeteer
  Headless Chrome automation from Google for scraping and testing. github.com/puppeteer/puppeteer
• Scrapy
  The most powerful Python web scraping framework, built for large-scale crawling. github.com/scrapy/scrapy
• Browserless
  Self-hosted headless-Chrome-as-a-service — drop-in replacement for cloud Selenium/Puppeteer. github.com/browserless/browserless
• ScrapeGraphAI
  Uses LLMs to scrape websites from natural-language instructions. github.com/ScrapeGraphAI/Scrapegraph-ai
• Jina Reader
  Converts any URL to clean markdown for LLMs with a single API call. github.com/jina-ai/reader
• Markitdown (Microsoft)
  Converts PDFs, Office docs, and images to Markdown for AI pipelines. github.com/microsoft/markitdown

🏠 10. Self-Hosted Apps & Media

• Jellyfin
  Open-source Netflix alternative — self-host your entire media library. github.com/jellyfin/jellyfin
• Immich
  Self-hosted Google Photos alternative with AI face recognition and mobile backup. github.com/immich-app/immich
• Home Assistant
  The ultimate smart-home automation platform — 3,000+ integrations, fully local. github.com/home-assistant/core
• Ntfy
  Simple push notifications via HTTP — alerts from any script to your phone. github.com/binwiederhier/ntfy
• Uptime Kuma
  Beautiful self-hosted uptime monitoring with alerting. github.com/louislam/uptime-kuma
• IT-Tools
  60+ developer utilities (encode, decode, convert, generate) in one clean self-hosted app. github.com/CorentinTh/it-tools
• Dashy
  Open-source personal dashboard for all your self-hosted services. github.com/Lissy93/dashy

🧠 11. Research, Music & Hidden Gems

• AudioX
  Unified framework integrating text, video, image, and audio conditions for multimodal audio synthesis. github.com/ZeyueT/AudioX
• Demucs (Meta)
  State-of-the-art music source separation — split any song into vocals, drums, bass, and instruments. github.com/facebookresearch/demucs
• Whisper (OpenAI)
  The original open speech recognition model, transcribing 99 languages with near-human accuracy. github.com/openai/whisper
• LLaVA
  Open large vision-language model — describe images, answer visual questions, analyze screenshots, fully local. github.com/haotian-liu/LLaVA
• MusicGen (Meta)
  Generate music from text prompts — open-source, self-hosted, no limits. github.com/facebookresearch/audiocraft
• awesome-ai-agents-2026
  177 production-ready SOUL.md configs across 24 categories (PM, SEO, DevOps, Writer, Support) for OpenClaw. github.com/caramaschiHG/awesome-ai-agents-2026

🗺️ Quick category map