New Trending AI repos on Github: September 2025
Trending Open-Source Projects — September 2025
This month’s trending repositories can be logically grouped into the following categories: 3D Vision & Reconstruction, AI Agent Frameworks & Educational Resources, AI-Powered Development & Debugging Tools, Data Management & Query Interfaces, and Cross-Platform Profiling & Performance Analysis.
3D Vision & Reconstruction
MapAnything
Stars: 1,681 | Created: September 4, 2025
A universal, feed-forward transformer model for metric 3D reconstruction supporting 12+ tasks including multi-view stereo, monocular depth, and registration. The model ingests flexible input combinations such as images, calibration data, depth, and poses, and outputs dense 3D geometry, camera parameters, and confidence maps. It is end-to-end trained, supports memory-efficient inference, and integrates with COLMAP and Gaussian Splatting through export tools. Available under two licenses: research (CC-BY-NC) and commercial (Apache 2.0).
AI Agent Frameworks & Educational Resources
Paper2Agent
Stars: 635 | Created: September 9, 2025
Transforms research papers into interactive AI agents by extracting executable tools from codebases and running them on MCP servers. Supports automated tutorial processing, isolated environments, and integrates with AI coding assistants like Claude Code. Agents can be executed locally or remotely (e.g., Hugging Face servers), facilitating natural language interaction with complex scientific codebases.
LLM-Agents-Ecosystem-Handbook
Stars: 333 | Created: September 8, 2025
An educational compendium featuring over 60 scaffolded projects focused on large language model (LLM) agents. Provides a comprehensive framework comparison matrix, practical tutorials on retrieval-augmented generation (RAG), memory, fine-tuning, and multi-agent systems with vendor-neutral guidance. Includes an agent skeleton generator and evaluation tools suitable for both beginners and advanced practitioners.
Agentic Design Patterns (Russian)
Stars: 322 | Created: September 4, 2025
A practical guide to building intelligent multi-agent systems, covering 21 design patterns ranging from prompt chaining and planning to inter-agent communication, security, and adaptability. Includes examples using LangChain, CrewAI, RAG, and MCP. All author royalties are donated to charity.
PAI: Personal AI Infrastructure
Stars: 305 | Created: September 8, 2025
Open-source modular AI infrastructure designed to augment personal and work life. Features dynamic hooks, custom commands, specialized AI agents, and MCP integrations across domains such as finance, health, content creation, and automation. Employs plaintext configuration to support diverse AI platforms, emphasizing extensibility, user ownership, and lifelong adaptability.
AI-Powered Development & Debugging Tools
ai-sdk-tools
Stars: 630 | Created: September 10, 2025
Extends the Vercel AI SDK with advanced features for state management, debugging, and structured streaming of AI-generated outputs. Includes modules like @ai-sdk-tools/store for scalable chat state, @ai-sdk-tools/devtools for enhanced debugging, and @ai-sdk-tools/artifacts for type-safe real-time streaming of content such as dashboards and documents.
dev3000
Stars: 416 | Created: September 2, 2025
Captures full-stack development activity—including server logs, browser console messages, network requests, and screenshots—into an AI-readable, timestamped chronological feed. Supports integration with AI tools via an MCP server for efficient debugging. Operates in two modes: default Playwright-controlled browser and lightweight Chrome extension. Logs are accessible through a local web interface.
Disciplined AI Software Development
Stars: 330 | Created: September 4, 2025
Introduces a disciplined, four-stage methodology for AI software projects that emphasizes structured planning, behavioral constraints, and empirical validation. Utilizes AI personas and custom instructions for consistency and limits file sizes for better context management. Supports benchmarking and iterative, data-driven development with architectural compliance verification.
Data Management & Natural Language Query Interfaces
PGMCP
Stars: 452 | Created: September 16, 2025
A PostgreSQL Model Context Protocol server enabling natural language queries over any PostgreSQL database through AI assistants. Supports MCP-compatible clients such as Cursor and Claude Desktop. Translates English queries to safe, read-only SQL with streaming support for large datasets. Features robust error handling, case-sensitive identifier support, multi-format outputs (table, JSON, CSV), and free-text column search—all without schema alterations.
Cross-Platform Profiling & Performance Analysis
uniprof
Stars: 343 | Created: September 10, 2025
A zero-configuration, unified CPU profiling tool supporting multiple platforms and runtimes, including Python, Node.js, Ruby, JVM, .NET, PHP, BEAM, and native Linux/macOS. Works both inside containers and on host systems by automatically selecting available profilers and normalizing outputs. Provides hotspot analysis and flamegraph visualizations without requiring application changes. Includes an MCP server interface for AI agent integration.
These projects demonstrate cutting-edge advances across AI model integration, developer tooling, natural language interface design, and system observability—underscoring the vibrant innovation in the AI open-source ecosystem this September.