Home
Curated open-source AI projects with verified GitHub data and hands-on breakdowns.
img2threejs/img2threejs (11,315 stars, Python, Apache-2.0, pushed yesterday) rebuilds the object in a reference image as a code-only procedural Three.js model -- token-efficient, animation-ready, reconstruction-by-code not photogrammetry/mesh. Pitfalls: output is code not mesh, needs Three.js runtime, strong for rule-based objects, weak for complex organic shapes.
code-yeongyu/oh-my-openagent (67k stars, TypeScript, pushed today) is a batteries-included OpenCode plugin with multi-model orchestration, parallel background agents, and custom LSP/AST tools. Its README declares "Anthropic blocked OpenCode because of us -- Claude Code is a nice prison." Pitfall: licensed under SUL-1.0, not MIT, with usage restrictions.
charmbracelet/crush (27,271 stars, Go, pushed today) is an open-source terminal coding agent, model-agnostic (OpenAI/Anthropic-compatible, works with GLM/DeepSeek). Pitfalls: GitHub reports NOASSERTION license (verify against LICENSE file), terminal-not-IDE, self-paid model API, MCP config. Differs from Claude Code/Cursor/openclaw.
trycua/cua (21,114 stars, MIT, pushed today) is an open-source computer-use 2.0 stack: background drivers (macOS/Windows/Linux, no cursor stealing), cross-OS sandboxes, Cua Bench benchmarks, and Lume virtualization. Unlike browser-use (browser only) or openclaw (email/calendar), cua drives the entire OS.
OpenMontage (46,548 stars, Python/AGPL-3.0) is an open-source agentic video production system whose defaults depend on foreign cloud models (OpenAI/Veo/Runway/ElevenLabs) blocked in China by VPN, payment, and phone-number barriers; but PROVIDERS.md natively supports Kling/Qwen/Jimeng/Doubao/DeepSeek and local Wan2.1 -- with a full domestic-alternative mapping table and minimal working .env config.
NousResearch/hermes-agent is an AI agent with a built-in learning loop - 227,993 stars, Python, MIT license, created 2025-07-22, pushed today. Core features: skill distillation from experience, self-improvement, knowledge persistence, past-conversation search, Honcho user modeling, 7 backends (incl. serverless sleep), model-agnostic, multi-platform messaging gateway.
citrolabs/ego-lite is an open-source Chromium-based browser built for human + agent sharing (MIT, JavaScript, 9,335 GitHub stars, #9 on Trending weekly). Core selling point: agents reuse your real login state and tabs via the ego-browser skill and run tasks in parallel in their own Spaces, differentiating it from automation frameworks like browser-use that spin up a clean separate browser. macOS only today; Windows/Linux on roadmap.
Ponytail is a skill that plugs into AI coding agents, making them climb a seven-rung laziness ladder before writing any code. Real Claude Code benchmarks show 54% less code, 20% cheaper, 27% faster, with every safety guard intact.
koala73/worldmonitor is an open-source real-time global intelligence dashboard written in TypeScript under AGPL v3, with 79,487 GitHub stars. It uses AI to aggregate 500+ news feeds, geopolitical signals, market data, and infrastructure status; a dual map engine (globe.gl + deck.gl) with 56 layers, cross-stream correlation of military/economic/disaster/escalation signals, a Country Instability Index scoring 31 Tier-1 countries, local Ollama inference with no API key, six site variants from one codebase, a Tauri 2 desktop app, and 26 languages with RTL support.
A 14,204-star Claude Code skill where a single /watch command downloads a video, extracts frames, transcribes it, and hands everything to Claude, so the AI finally sees the picture instead of only reading captions. #20 on GitHub weekly trending, MIT, Python.