Today's AI Frameworks & SDKs: Fastest-Growing Projects — May 15, 2026
Today's the AI Frameworks & SDKs space, there's a noticeable trend towards community-driven projects that support various large language models (LLMs) and facilitate interactions with these models across different hardware platforms. One such project is MaaNTE, which stands out for its strong growth score despite having two distinct GitHub repositories. Another emerging project to watch is club-3090, a repository providing recipes for running LLMs on RTX 3090 GPUs, demonstrating significant activity and community engagement.
1bananachicken/MaaNTE: This repository houses an automation assistant called MaaNTE, which supports features like auto-fishing, auto-selling fish, and auto-buying bait. It appears to be growing rapidly, with a high growth score of 58.24 and over 1,000 stars, likely due to its unique set of automated features tailored for specific gaming environments.
TheRunicDev/MaaNTE: Similar to the previous entry, this repository also focuses on MaaNTE but provides additional details about features such as auto-coffee making and cafe revenue extraction. Despite fewer recent commits, it has a solid growth score of 51.33 with nearly 300 stars, indicating sustained interest from users looking for automation tools in gaming.
noonghunna/club-3090: This project offers community recipes to run LLMs on RTX 3090 GPUs using multiple engines like vLLM and llama.cpp. With a growth score of 48.38 and over 900 stars, its rapid development (100 commits in the last month) is driving significant interest from developers seeking efficient ways to deploy large models on powerful hardware.
alash3al/stash: Stash provides a persistent memory layer for AI agents using PostgreSQL, offering an MCP server. It has garnered 692 stars and a growth score of 32.48, suggesting strong community engagement and active development, as evidenced by nearly 100 commits in the past month.
wanghuan9/skill-manager: SkillDock is a desktop control center that manages AI skills and MCP servers with Git-aware updates for tools like Claude Code and GitHub Copilot. With only 55 stars but a respectable growth score of 31.75, it's growing steadily due to its unique approach in managing various AI integrations.
oritera/Cairn: Cairn is described as a state-space search engine validated through autonomous penetration testing. It has amassed over 900 stars and a growth score of 30.52, likely due to its versatile application across different domains requiring complex problem-solving capabilities.
iFurySt/open-codex-computer-use: This project offers an open-source alternative to Codex Computer Use, aiming to be accessible for broader use cases. With nearly 800 stars and a growth score of 29.05, it's growing steadily due to its community-driven nature and the need for more accessible alternatives in AI tooling.
Ontos-AI/knowhere: Knowhere extracts and parses structured data ready for AI agents and retrieval-augmented generation (RAG) systems. It has attracted over 147 stars and a growth score of 26.60, reflecting active development efforts with 100 commits in the last month.
ZhiYi-R/moon-bridge: This project acts as a forwarding layer to convert messages from Anthropic into OpenAI-compatible responses for Codex. With over 163 stars and a growth score of 20.95, it's growing due to its utility in bridging different AI service ecosystems.
TheTom/vllm-swift: vLLM Swift is a high-performance LLM inference plugin for Apple Silicon devices using mlx-swift technology. It has gained over 259 stars and a growth score of 20.68, highlighting the interest in optimizing large language models for newer hardware platforms.
Each project offers unique features tailored to specific needs within the AI ecosystem, from automation tools for gaming environments to optimized LLM inference on modern GPUs and Apple Silicon devices. The high levels of activity and community engagement underscore the dynamic nature of development in this space.
1bananachicken/MaaNTE: This repository houses an automation assistant called MaaNTE, which supports features like auto-fishing, auto-selling fish, and auto-buying bait. It appears to be growing rapidly, with a high growth score of 58.24 and over 1,000 stars, likely due to its unique set of automated features tailored for specific gaming environments.
TheRunicDev/MaaNTE: Similar to the previous entry, this repository also focuses on MaaNTE but provides additional details about features such as auto-coffee making and cafe revenue extraction. Despite fewer recent commits, it has a solid growth score of 51.33 with nearly 300 stars, indicating sustained interest from users looking for automation tools in gaming.
noonghunna/club-3090: This project offers community recipes to run LLMs on RTX 3090 GPUs using multiple engines like vLLM and llama.cpp. With a growth score of 48.38 and over 900 stars, its rapid development (100 commits in the last month) is driving significant interest from developers seeking efficient ways to deploy large models on powerful hardware.
alash3al/stash: Stash provides a persistent memory layer for AI agents using PostgreSQL, offering an MCP server. It has garnered 692 stars and a growth score of 32.48, suggesting strong community engagement and active development, as evidenced by nearly 100 commits in the past month.
wanghuan9/skill-manager: SkillDock is a desktop control center that manages AI skills and MCP servers with Git-aware updates for tools like Claude Code and GitHub Copilot. With only 55 stars but a respectable growth score of 31.75, it's growing steadily due to its unique approach in managing various AI integrations.
oritera/Cairn: Cairn is described as a state-space search engine validated through autonomous penetration testing. It has amassed over 900 stars and a growth score of 30.52, likely due to its versatile application across different domains requiring complex problem-solving capabilities.
iFurySt/open-codex-computer-use: This project offers an open-source alternative to Codex Computer Use, aiming to be accessible for broader use cases. With nearly 800 stars and a growth score of 29.05, it's growing steadily due to its community-driven nature and the need for more accessible alternatives in AI tooling.
Ontos-AI/knowhere: Knowhere extracts and parses structured data ready for AI agents and retrieval-augmented generation (RAG) systems. It has attracted over 147 stars and a growth score of 26.60, reflecting active development efforts with 100 commits in the last month.
ZhiYi-R/moon-bridge: This project acts as a forwarding layer to convert messages from Anthropic into OpenAI-compatible responses for Codex. With over 163 stars and a growth score of 20.95, it's growing due to its utility in bridging different AI service ecosystems.
TheTom/vllm-swift: vLLM Swift is a high-performance LLM inference plugin for Apple Silicon devices using mlx-swift technology. It has gained over 259 stars and a growth score of 20.68, highlighting the interest in optimizing large language models for newer hardware platforms.
Each project offers unique features tailored to specific needs within the AI ecosystem, from automation tools for gaming environments to optimized LLM inference on modern GPUs and Apple Silicon devices. The high levels of activity and community engagement underscore the dynamic nature of development in this space.