Trending AI projects

A framework for distilling an author's writing style into a structured "Writing DNA" that AI agents can use to stably and accurately replicate that style.

AG‑UI is an open, lightweight, event‑based protocol that standardises how AI agents interact with user‑facing applications. It defines ~16 event types, supports any transport (SSE, WebSocket, webhook), and provides middleware, SDKs (Kotlin, Go, Rust, etc.) and first‑party integrations (LangChain, CrewAI, Microsoft Agent Framework, CopilotKit). Features include real‑time streaming chat, bi‑directional state sync, generative UI messages, and human‑in‑the‑loop collaboration. The project offers a CLI starter, a Dojo of small examples, and is MIT‑licensed with an active community.

A self-hosted gateway that bridges Qoder domestic and international accounts into standard OpenAI and Anthropic-compatible APIs, featuring multi-account management and request routing.

A video creation CLI that enables AI Agents to assemble editable video projects from raw footage and scripts, allowing creators to automate rough cuts and packaging while retaining manual refinement capabilities.

nono is a lightweight, cross‑platform runtime that sandbox‑isolates AI agents and the tools they call. It enforces per‑tool file, network and credential policies, verifies assets with Sigstore, logs tamper‑evident audits, and supports rollback. Installable via a single binary (curl, Homebrew, Nix, etc.), it runs unprivileged on laptops, containers or CI, and offers Rust/Python bindings for integration.

OpenCollab is a Python framework for orchestrating multiple coding agents (analyst, coder, tester, etc.) via configurable YAML teams and programmable Python workflows. It enforces the declared collaboration graph, records detailed traces, and measures *Adherence*—how often agents follow the plan. Benchmarks show it outperforms larger single‑agent code generation systems on SWE‑bench, Terminal‑Bench, and DeepSWE, while remaining under 50 k lines of code.

CodexBar is a menu bar utility for macOS and Linux that tracks usage limits, billing, and reset countdowns across over 100 AI coding providers and LLM services.

A collection of modular video-editing skills for AI agents that automate tasks like motion-graphic B-roll generation, object separation, and product explainer creation.

assistant‑ui is a TypeScript/React library that provides composable, production‑ready chat UI components for AI assistants. It works with many LLM backends via typed runtimes, handles streaming, markdown, tool‑call rendering, and offers a CLI to scaffold or inject the UI into Next.js projects. MIT‑licensed and backed by Y Combinator.

A private AI-powered movie library manager that automates metadata scraping, file organization, and poster enhancement using a local-first approach.

Graph-Native Infrastructure for Context and Accountable AI Systems

pi‑subagents is a Pi AI‑assistant extension that lets the main Pi session spawn focused child agents (scout, reviewer, worker, oracle, etc.) to handle tasks like code review, research, implementation, and auditing. Subagents can run in the foreground (streaming live) or in the background (detached process). Users simply install the package (`pi install npm:pi-subagents`) and then ask Pi in plain language (e.g., “Use reviewer to review this diff”). The system orchestrates the child agents, provides observability via a FleetView panel, and offers configuration options such as tool activation and spawn limits.

A lightweight library for building AI agents that can execute tasks by writing Python code, offering model and tool agnostic integrations with sandboxed execution support.

A persistent, project-aware memory layer for AI coding agents that prevents them from forgetting project context across sessions using a local SQLite database and MCP.

A production-ready MCP server that integrates Ghidra with AI agents, providing over 200 tools for automated binary analysis, reverse engineering, and consistent documentation.

PR‑Agent is an open‑source AI bot that automatically reviews pull‑requests (or issues) using LLMs. It can summarize changes, give a full code‑review, suggest improvements, or answer free‑form questions. The tool works with GitHub, GitLab, Bitbucket, Azure DevOps, and Gitea, and can be run as a GitHub Action, Docker container, or local CLI. Prompts are fully configurable, and it supports many LLM providers via LiteLLM. Self‑hosting keeps data private, and the project is community‑maintained with sponsorship from Qodo.

A comprehensive toolkit for creating professional scientific data plots and system architecture diagrams, featuring AI-assisted generation, reusable SVG assets, and native PowerPoint export capabilities.

A plugin for DeepSeek Harness that enables free, multi-engine web searching with automatic fallback, removing the need for a mandatory official API key.

CC GUI is an IntelliJ IDEA plugin that provides a chat‑style interface for several AI code‑generation models (Claude Code, OpenAI Codex, etc.). It supports file and image context, diff viewing, session history, and a slash‑command‑driven agent system, letting developers interact with AI assistants directly inside their IDE.

PenguinHarness is an open‑source, cross‑platform framework for building, benchmarking, and automatically improving AI agents (RSI). It ships with multiple self‑evolution algorithms, a suite of benchmarks, and a UI that can generate full‑stack apps from a one‑sentence description. Supports many LLM providers, offers desktop, web, CLI, Docker, and Chrome‑extension interfaces, and is released under Apache‑2.0.

Ming Voice is a fully local, real-time voice assistant that runs on your GPU to provide private, low-latency control over smart home devices and personal tasks without sending audio to the cloud.

MuJoCo is a general-purpose physics engine from Google DeepMind for simulating articulated structures that interact with their environment. It provides a high-performance C API, Python bindings, and interactive 3D visualization, making it a popular tool in robotics and machine learning research.

An AI Agent Skill that distills the complete TCM thinking system of Ni Haixia, enabling AI to perform Six Meridian syndrome differentiation and prescription selection based on 3.5M words of distilled knowledge.

A DeepSeek Harness plugin that enables users and AI to search, send, and learn memes, supporting both manual selection and automated AI-driven meme battles across web and IM platforms.

SANA is an efficiency-focused framework for high-resolution image and video generation that provides optimized training and inference pipelines to significantly reduce compute requirements and latency.

An agent skill that upgrades AI-generated motion graphics by applying professional design rules, measured timing, and automated quality checks to eliminate common AI-generated aesthetic flaws.

A collection of 26 professional screenwriting and dramaturgy skills for AI agents, distilled from dozens of craft books and published scripts to assist with structure, dialogue, and series development.

An MCP server that lets AI assistants control DaVinci Resolve video editing software through its official scripting API, with full editing, grading, and rendering capabilities, plus an offline file-level server for cloud workflows.

A portable writing skill for AI agents that audits and rewrites content to remove formulaic AI writing patterns and "AI-isms" using a tiered vocabulary and pattern catalog.

An open-source product integration platform that provides a unified REST API syntax for connecting AI agents and backend services to multiple third-party apps.

dsh‑mnemon is a DeepSeek Harness plugin that adds layered, visible memory (runtime, project docs, long‑term spaces) and self‑evolving skill management, with a UI for inspecting turn‑level reads/writes and an SDK for custom sources, strategies, and storage providers.

A set of AI agent skills that enable end-to-end searching, comparing, and booking of worldwide hotels and flights through natural language conversations.

A self‑hosted MCP server that lets AI assistants read/write Xiaohongshu (login, post images/video, search, comment, like, fetch profiles, etc.). Install via binary, Docker, or source, log in once, start the service on port 18060, and connect any MCP‑compatible client (Claude Code, Cursor, VS Code, Open Claw, etc.).

A collection of AI-integrated market research tools for analyzing stocks, options, and digital assets, providing financial data and strategy analysis within an AI workspace.

Keystone Bench is a Python library and data set that evaluates medical chat assistants on how they adapt when a single decisive fact in a patient‑doctor conversation is removed, contradicted, or altered. It builds on OpenAI’s HealthBench, adds 7,419 edited “twins” across nine families, provides clinician‑written rubrics, and uses a separate LLM judge to label model replies as acceptable or forbidden. The benchmark offers layered splits, reproducible reference results, and a full audit pipeline, enabling researchers to measure evidence‑sensitivity and robustness of AI assistants in clinical settings.

Browser Harness is an open‑source Python bridge that lets LLM coding agents control a real Chrome browser via the DevTools Protocol, enabling logged‑in, UI‑driven automation that improves as agents write reusable helper code.

A secure, isolated sandbox platform using Firecracker microVMs that allows AI agents and applications to safely execute code and manage environment lifecycles via an HTTP API.

An MCP server that turns AI assistants into security analysts by integrating 28 security tools and 24 APIs to automate CVE triage and risk scoring.

A cost-aware workbench for DeepSeek Harness Desktop that enables users to plan, route, and execute complex tasks across AI models and official tools while managing budgets and permissions.

A npm‑based CLI and web catalog that supplies ready‑made Claude Code agents, commands, settings, hooks, and external‑service integrations (MCPs). Install components with one‑line `npx claude-code-templates@latest …` commands, view analytics, monitor chats, and run health checks. MIT‑licensed, aggregating community and Anthropic‑official templates.

Phocinae-Largha-150M-v1 is a compact, bilingual decision-making model that provides deterministic, local, and low-latency answers for structured tasks like command approval and tool routing without generating text.

An offline transcription and translation tool powered by OpenAI's Whisper that allows users to convert audio and video to text locally on their own hardware.

Hermes Jev Skills is a Python‑only plug‑in suite that lets the Hermes (or Claude Code / Codex) AI‑agent off‑load cheap, fast decision‑making (model routing, search ranking, mailbox sorting, turn selection, etc.) to TypeSafe’s Jev decision model. The repo provides skill files, a `jev` CLI, a dashboard, and an installer that integrates with existing agents. All data is redacted before leaving the machine, and the system fails open if Jev is unavailable.

Sphinx adds an understanding check to agent-assisted development: it pauses a git push and asks you to predict what your change will do, explaining mistakes and re-quizzing until you pass. It works with coding agents like Claude Code and Codex, using a reviewer model to generate open-book questions about behavior and failure modes.

Gas Town is an open‑source Go‑based workspace manager that lets you run many AI coding agents (Claude, Copilot, Codex, Gemini, etc.) on multiple Git projects while persisting every step in a Git‑backed ledger. A central AI coordinator (the *Mayor*) creates *convoys* of work items (*beads*), hands them to worker agents (*Polecats*), and a set of watchdog services (Witness, Deacon, Dogs) keep the system healthy. Completed work is merged via a Bors‑style queue (*Refinery*), and an optional federation layer (*Wasteland*) lets different installations share tasks. Installable via Homebrew, `go install`, or Docker, it provides a CLI (`gt`) for creating rigs, crews, convoys, and for monitoring progress.

A bridge connecting MCP-compatible AI assistants to the Godot editor, enabling AI to build scenes, edit nodes, and write scripts directly in the engine.

An open-source, voice-interactive AI companion featuring a Live2D avatar and visual perception, capable of running entirely offline for private, real-time conversations.

Short Video Factory is an open‑source desktop tool that uses AI (OpenAI‑style text generation and Edge‑TTS voice synthesis) to turn a short prompt and raw footage into a finished short video, with subtitles, batch processing and cross‑platform support.