Hot open-source projects on GitHub — 97 articles
MarkTechPost · May 23
Tencent open-sourced TencentDB Agent Memory, a fully local 4-tier memory system for AI agents that replaces flat vector stores with a semantic pyramid (Conversation → Atom → Scenario → Persona). Ships
GitHub Trending (Dev Tools) · May 21
DeepSeek-Reasonix is a terminal coding agent engineered specifically around DeepSeek's prefix-cache mechanics to keep token costs low across multi-hour sessions. The design treats cache stability as a
GitHub Trending (Dev Tools) · May 21
GitHub's spec-kit is an open-source toolkit that makes specifications executable—they generate working code directly instead of just guiding implementation. Install via `uv tool install specify-cli`,
GitHub Trending · May 19
ECC is a performance optimization system for AI coding agents (Claude Code, Cursor, GitHub Copilot) that bundles skills, memory optimization, security scanning, and agent configuration templates. Buil
GitHub Trending (Dev Tools) · May 18
DESIGN.md—a markdown-based design system format that AI agents read natively—lets you drop a style guide into your project and have Claude/Copilot generate pixel-matched UI without Figma exports or JS
MarkTechPost · May 14
Cline extracted its internal agent harness into @cline/sdk, an open-source TypeScript SDK now powering its CLI, Kanban, and IDE extensions. The four-layer stack (shared, LLMs, agents, core) supports p
NVIDIA AI · May 13
Hermes Agent, an open-source agentic framework from Nous Research, hit 140,000 GitHub stars in three months and is now the most-used agent globally on OpenRouter. It features self-evolving skills, con
Claude Code · May 6
What's changed Fixed VS Code extension failing to activate on Windows due to a hardcoded build path in the bundled SDK (createRequire polyfill bug) Fixed Mantle endpoint authentication failing with mi
Analytics Vidhya · May 3
A curated roundup of 15+ open-source agentic AI projects with working code—covering agent architectures, tool use, multi-step reasoning, and autonomous workflows that practitioners can fork and study.
Lobsters AI · May 2
Fabrica is a lightweight terminal-based harness for running coding agents with LLM backends. The project prioritizes simplicity and local control over heavy frameworks, supporting Claude, Ollama, and
IEEE Spectrum · Apr 30
DAIMON Robotics released Daimon-Infinity, a multimodal dataset with 10,000+ hours of high-resolution tactile sensing data for robot manipulation tasks, built on their 110,000-unit fingertip sensor and
GitHub Trending · Apr 27
Browser Harness lets LLMs control your browser and fix themselves on the fly: if an agent needs a helper function to complete a task, it writes the code directly into helpers.py mid-execution. Minimal
GitHub Trending · Apr 27
Open-multi-agent converts high-level goals into optimized task DAGs and executes them in parallel across multiple agents and models—TypeScript-native with just 3 runtime dependencies. Engineers descri