Today's AI Frameworks & SDKs: Fastest-Growing Projects — May 18, 2026
Today's the AI Frameworks & SDKs space, there's a noticeable trend towards high-performance and multi-agent systems designed to enhance efficiency across various domains such as trading, orchestration, and data handling. Among these tools, several projects stand out for their unique approaches and strong community engagement, with TokenSpeed leading the pack due to its cutting-edge inference engine capabilities.
TokenSpeed, developed by lightseekorg, is a speed-of-light LLM inference engine designed to optimize performance in real-time applications. Its impressive growth score of 79.92 and over 1,000 stars on GitHub reflect its popularity among developers seeking fast and efficient large language model deployments.
TradingAgents-astock, created by simonlin1212, is a multi-agent investment research framework tailored specifically for the A-share market in China, offering sophisticated risk assessment tools. The project’s steady growth score of 64.40 and growing number of stars indicate its increasing relevance among financial analysts and traders looking to leverage AI-driven insights.
Harmonist-Orchestral by 2508965-ship-it is a multi-agent orchestration engine aimed at building and deploying AI swarms using Claude Code, showcasing future-oriented capabilities in automation and coordination. With a growth score of 53.25, the project demonstrates strong traction among developers interested in advanced agent-based systems.
MaaNTE by 1bananachicken is an assistant driven by MAAFramework designed to automate tasks within specific gaming environments, with features such as auto-fishing and revenue extraction. Its high growth score of 52.61 and over 1,300 stars highlight its appeal among gamers seeking automation solutions for repetitive tasks.
Another version of MaaNTE, developed by TheRunicDev, is an automation assistant designed for fishing, selling fish, and buying bait in specific games, with additional features like auto-coffee making. Despite a slightly lower growth score of 44.42, the project's growing star count suggests it is gaining traction among users who require comprehensive automation tools.
Club-3090 by noonghunna provides community recipes for serving large language models on RTX 3090 GPUs, supporting multiple engines and models with specific configurations for Qwen3.6-27B. With a growth score of 43.15, the project is gaining popularity among developers looking to optimize model performance across different hardware setups.
Cairn by oritera is a general-purpose state-space search engine initially validated through autonomous penetration testing, showcasing its versatility in various applications requiring complex search algorithms. Its steady growth score of 30.29 and over 1,000 stars indicate its growing importance among developers working on AI-driven security solutions.
Stash by alash3al is a persistent memory layer for AI agents that stores episodes, facts, and working context in Postgres, offering an MCP server and self-hosting capabilities without cloud requirements. With a growth score of 28.52, the project demonstrates strong interest among developers seeking robust storage solutions for AI applications.
Knowhere by Ontos-AI is a tool designed to extract, parse, and output structured chunks ready for use with AI agents and retrieval-augmented generation models. Its growing star count and growth score of 25.39 suggest its relevance in the data processing and agent-oriented development communities.
Finally, OpenFoundry by DioCrafts offers an open-source alternative to Palantir Foundry, providing a platform for connecting any data source, building ontologies, creating pipelines, visualizing with dashboards, and making AI-powered decisions. With a growth score of 21.12 and increasing star count, the project is gaining attention from developers looking for self-hosted solutions in enterprise-level data management.
These projects highlight the diversity and innovation within the AI frameworks and SDKs space, catering to both specialized needs and broader applications across various industries.
TokenSpeed, developed by lightseekorg, is a speed-of-light LLM inference engine designed to optimize performance in real-time applications. Its impressive growth score of 79.92 and over 1,000 stars on GitHub reflect its popularity among developers seeking fast and efficient large language model deployments.
TradingAgents-astock, created by simonlin1212, is a multi-agent investment research framework tailored specifically for the A-share market in China, offering sophisticated risk assessment tools. The project’s steady growth score of 64.40 and growing number of stars indicate its increasing relevance among financial analysts and traders looking to leverage AI-driven insights.
Harmonist-Orchestral by 2508965-ship-it is a multi-agent orchestration engine aimed at building and deploying AI swarms using Claude Code, showcasing future-oriented capabilities in automation and coordination. With a growth score of 53.25, the project demonstrates strong traction among developers interested in advanced agent-based systems.
MaaNTE by 1bananachicken is an assistant driven by MAAFramework designed to automate tasks within specific gaming environments, with features such as auto-fishing and revenue extraction. Its high growth score of 52.61 and over 1,300 stars highlight its appeal among gamers seeking automation solutions for repetitive tasks.
Another version of MaaNTE, developed by TheRunicDev, is an automation assistant designed for fishing, selling fish, and buying bait in specific games, with additional features like auto-coffee making. Despite a slightly lower growth score of 44.42, the project's growing star count suggests it is gaining traction among users who require comprehensive automation tools.
Club-3090 by noonghunna provides community recipes for serving large language models on RTX 3090 GPUs, supporting multiple engines and models with specific configurations for Qwen3.6-27B. With a growth score of 43.15, the project is gaining popularity among developers looking to optimize model performance across different hardware setups.
Cairn by oritera is a general-purpose state-space search engine initially validated through autonomous penetration testing, showcasing its versatility in various applications requiring complex search algorithms. Its steady growth score of 30.29 and over 1,000 stars indicate its growing importance among developers working on AI-driven security solutions.
Stash by alash3al is a persistent memory layer for AI agents that stores episodes, facts, and working context in Postgres, offering an MCP server and self-hosting capabilities without cloud requirements. With a growth score of 28.52, the project demonstrates strong interest among developers seeking robust storage solutions for AI applications.
Knowhere by Ontos-AI is a tool designed to extract, parse, and output structured chunks ready for use with AI agents and retrieval-augmented generation models. Its growing star count and growth score of 25.39 suggest its relevance in the data processing and agent-oriented development communities.
Finally, OpenFoundry by DioCrafts offers an open-source alternative to Palantir Foundry, providing a platform for connecting any data source, building ontologies, creating pipelines, visualizing with dashboards, and making AI-powered decisions. With a growth score of 21.12 and increasing star count, the project is gaining attention from developers looking for self-hosted solutions in enterprise-level data management.
These projects highlight the diversity and innovation within the AI frameworks and SDKs space, catering to both specialized needs and broader applications across various industries.