Today's AI Frameworks & SDKs: Fastest-Growing Projects — April 16, 2026
Today's AI Frameworks & SDKs space saw a surge in growth, driven by innovative solutions for AI agent management, multi-modal modeling, and language model access to formal verification. The top-growing repositories showcase a mix of established players and newcomers, all vying for attention with their unique value propositions. As the AI landscape continues to evolve, these frameworks and SDKs are poised to play a crucial role in shaping its future.
Rohitg00's "ai-engineering-from-scratch" repository takes the top spot with a growth score of 89.90 and over 2,838 stars. This comprehensive resource allows developers to learn, build, and ship AI solutions from scratch, making it an attractive destination for those looking to get started with AI engineering. Its massive popularity stems from its ability to provide a structured approach to building AI systems.
Jxnxts' "mcp-brasil" repository comes in second, boasting a growth score of 59.71 and 1,374 stars. This MCP server provides access to 41 public APIs in Brazil, making it an essential tool for developers working with Brazilian data sources. Its growth can be attributed to the increasing demand for localized AI solutions.
Yogthos' "chiasmus" repository, with a growth score of 34.06 and 58 stars, offers an MCP server that grants language models access to formal verification. This unique feature makes it an attractive choice for developers working on safety-critical AI applications. Its growth is driven by the need for more robust and reliable AI systems.
The "ClawManager" repository from Yuan-lab-LLM has a growth score of 25.65 and 563 stars, providing a Kubernetes-native control plane for managing AI agent instances. Its governed AI access, runtime orchestration, and reusable resources make it an appealing solution for large-scale AI deployments. The growth of ClawManager can be attributed to the increasing adoption of Kubernetes in AI environments.
Rcortx's "kiwiq" repository, with a growth score of 24.50 and 1,058 stars, offers a production-grade multi-agent orchestration platform that has been battle-tested on over 200 enterprise AI agents. Its JSON-defined agents, multi-tier memory, and built-in observability make it an attractive choice for large-scale AI deployments. Despite having only three commits in the last 30 days, kiwiq's growth can be attributed to its established reputation as a reliable solution.
Codefromkarl's "ContextAtlas" repository has a growth score of 24.12 and 24 stars, providing context infrastructure for AI coding agents through hybrid retrieval, project memory, and retrieval observability. Its tree-sitter indexing, LanceDB vector search, FTS5, and token-aware context packing make it an appealing solution for developers working on AI-powered development tools. The growth of ContextAtlas can be attributed to the increasing demand for more efficient AI-powered coding solutions.
OpenEnvision's "Awesome-Multimodal-Modeling" repository boasts a growth score of 21.77 and 232 stars, offering a curated list of resources for multimodal modeling. Its comprehensive coverage of MLLM, UMM, and NMM makes it an essential resource for developers working on multimodal AI applications. The growth of this repository can be attributed to the increasing interest in multimodal AI research.
Other notable mentions include zju3dv's "habitat-gs" repository, which offers a high-fidelity navigation simulator with dynamic Gaussian splatting; rhino-acoustic's "NeuronFS", which provides an OS-native constraint engine for LLM agents; and jnMetaCode's "agency-orchestrator", which offers a multi-agent framework that works with existing AI subscriptions.
Rohitg00's "ai-engineering-from-scratch" repository takes the top spot with a growth score of 89.90 and over 2,838 stars. This comprehensive resource allows developers to learn, build, and ship AI solutions from scratch, making it an attractive destination for those looking to get started with AI engineering. Its massive popularity stems from its ability to provide a structured approach to building AI systems.
Jxnxts' "mcp-brasil" repository comes in second, boasting a growth score of 59.71 and 1,374 stars. This MCP server provides access to 41 public APIs in Brazil, making it an essential tool for developers working with Brazilian data sources. Its growth can be attributed to the increasing demand for localized AI solutions.
Yogthos' "chiasmus" repository, with a growth score of 34.06 and 58 stars, offers an MCP server that grants language models access to formal verification. This unique feature makes it an attractive choice for developers working on safety-critical AI applications. Its growth is driven by the need for more robust and reliable AI systems.
The "ClawManager" repository from Yuan-lab-LLM has a growth score of 25.65 and 563 stars, providing a Kubernetes-native control plane for managing AI agent instances. Its governed AI access, runtime orchestration, and reusable resources make it an appealing solution for large-scale AI deployments. The growth of ClawManager can be attributed to the increasing adoption of Kubernetes in AI environments.
Rcortx's "kiwiq" repository, with a growth score of 24.50 and 1,058 stars, offers a production-grade multi-agent orchestration platform that has been battle-tested on over 200 enterprise AI agents. Its JSON-defined agents, multi-tier memory, and built-in observability make it an attractive choice for large-scale AI deployments. Despite having only three commits in the last 30 days, kiwiq's growth can be attributed to its established reputation as a reliable solution.
Codefromkarl's "ContextAtlas" repository has a growth score of 24.12 and 24 stars, providing context infrastructure for AI coding agents through hybrid retrieval, project memory, and retrieval observability. Its tree-sitter indexing, LanceDB vector search, FTS5, and token-aware context packing make it an appealing solution for developers working on AI-powered development tools. The growth of ContextAtlas can be attributed to the increasing demand for more efficient AI-powered coding solutions.
OpenEnvision's "Awesome-Multimodal-Modeling" repository boasts a growth score of 21.77 and 232 stars, offering a curated list of resources for multimodal modeling. Its comprehensive coverage of MLLM, UMM, and NMM makes it an essential resource for developers working on multimodal AI applications. The growth of this repository can be attributed to the increasing interest in multimodal AI research.
Other notable mentions include zju3dv's "habitat-gs" repository, which offers a high-fidelity navigation simulator with dynamic Gaussian splatting; rhino-acoustic's "NeuronFS", which provides an OS-native constraint engine for LLM agents; and jnMetaCode's "agency-orchestrator", which offers a multi-agent framework that works with existing AI subscriptions.