Trending AI projects

Graphiti is an open-source framework for building temporal knowledge graphs that track how facts and relationships change over time, enabling AI agents to maintain evolving, context-aware memory.

Atticus-mac is a free, open-source voice assistant for macOS that lets you control your Mac by speaking: it can move windows, find files, read and click on screen content, dictate text, run background agents, and more. It works with Claude, Devin, Pi, or a local model, and keeps your voice private with on-device speech.

A local interface patch for Claude Desktop that adds support for Simplified and Traditional Chinese across macOS, Windows, and Windows.

Splash is a high-performance local inference engine for Apple Silicon that accelerates LLM execution using specialized Metal kernels and speculative decoding, providing an OpenAI-compatible API for agents and chat applications.

OpenResearch is a cross‑platform desktop and CLI app that lets AI coding agents (Claude, Codex, local LLMs, etc.) read scientific literature, generate hypotheses, run experiments on local or remote compute, and store every run in a local SQL database, all while keeping your data private.

Lody is an open‑source, cross‑platform workspace that lets teams share and coordinate AI coding agents (Claude Code, Codex, Kimi, etc.) across desktop, mobile, web and CLI. It syncs conversations, Git worktrees, files and visual feedback using CRDTs, enabling collaborative, version‑controlled AI‑driven development.

llmfit is a cross‑platform CLI/TUI (and optional web/API) tool that auto‑detects your CPU, RAM, GPU and accelerator setup, then ranks hundreds of open‑source LLMs (including various quantizations) by how well they fit, their estimated speed, quality and context length. It lets you benchmark models on your own hardware, share results, and consume recommendations via JSON or a REST API, making it easy to pick and deploy a model that actually runs on your machine.

LangAlpha is an open-source agentic framework for financial research and automated trading that maintains persistent investment theses and executes governed trades based on user-defined market signals.

An open format for creating AGENTS.md files to provide AI coding agents with project-specific context and instructions.

Token Optimizer is a zero‑dependency Python tool that plugs into AI coding assistants (Claude Code, Copilot, etc.) to automatically compress command output, search results, JSON, file diffs and other “structural” data, saving tokens and money. It adds smart checkpoints so savings survive model compaction, scores context quality live, stores a local SQLite audit trail, and provides slash‑commands (e.g., `/token‑optimizer`, `/token‑coach`) for reports and dashboards. Installation is a one‑time plugin add, after which it runs silently in the background.

PGX is a Postgres-compatible runtime that keeps hundreds or thousands of disposable databases alive in one process, with each database costing ~70 KB when idle and being created in milliseconds from a copy-on-write template — aimed at agent, CI, test, and preview workloads, and explicitly not a production primary (v0.1 is not crash-safe, no C extensions, single process).

Remory is a plug‑in that adds a learned “residual memory” to large language models, enabling them to compact long conversation histories while preserving important information. It works with frozen base models (Qwen 3.8‑27B or GLM‑5.3‑Flash) and provides a SGLang server, a Codex UI integration, and a Python API for summarising, encoding residual memory, and continuing generation from a compacted checkpoint.

A plugin for DeepSeek Harness that injects custom system prompts to bypass model safety filters, enabling unrestricted output for reverse engineering, cybersecurity, and creative tasks.

Claudemy is a Claude Code skill that generates personalized, bite-sized audio courses from current web research, designed to be practical and easy to finish.

Open-source platform that unifies LLMs (OpenAI, DeepSeek, local Ollama, etc.) into AI agents deployable across Slack, Discord, Telegram, WeChat, QQ, and more—with a web dashboard, plugin marketplace, and MCP/CLI tools so human or AI developers can manage every bot from one place.

OpenDataLoader PDF is an open‑source PDF parser that converts PDFs (including scanned ones) into Markdown, JSON with element bounding boxes, HTML, and auto‑tagged PDFs. It offers a fast deterministic Java‑based local engine plus an optional hybrid AI mode for complex tables, OCR, formula extraction, and image description, achieving top benchmark scores. The free core (Apache‑2.0) handles layout analysis and auto‑tagging; enterprise add‑ons provide full PDF/UA compliance and a visual accessibility studio.

Needle is a 8‑29 MB foundation model for edge devices that can do tool‑calling, structured JSON extraction, text embedding and speech‑to‑text (via the Whistle model). It uses a laddered Simple Attention architecture where every depth is a runnable sub‑model, and it ships as a single `.cact` binary with a custom C++ runtime. The Python SDK lets you decorate functions as tools, run the model locally, fine‑tune with LoRA or full‑model training, and deploy to many platforms.

Mastra is a TypeScript framework for building, orchestrating, and scaling AI agents and workflows with built-in support for model routing, human-in-the-loop processes, and production observability.

A framework using a standardized .motion file format that allows AI coding assistants to automatically generate professional motion-graphic videos from user scripts and source materials.

BeefTV is a desktop‑first, AI‑native video creation workspace. It provides a free‑form, node‑based canvas where users can connect prompts, assets and any text‑to‑image, video or audio model to generate and edit multimedia. Built with React, Go and Wails, it runs on macOS and Windows, stores data locally, and is extensible and lightweight.

Draw Agent Studio is a local, chat-driven drawing app: you describe or refine images in conversation, can also draw by hand, and export to common formats. It needs Python 3.11+, Node.js 20.19+, and a connected AI model with image input and tool-calling support.

A collection of first‑party plugins for the Hermes AI agent, providing integrations (Home Assistant, Spotify, Claude, Telegram, Blender, etc.) that can be installed individually via the Hermes CLI. Each plugin lives in its own directory with a manifest, tests, and optional MCP server, enabling lean core installs and rapid updates.

Slotstream is a native Swift engine for Apple Silicon that runs the Qwen3.8-Flash-Next mixture-of-experts model (125B parameters) on Macs with 16–64 GB memory by keeping most weights on SSD and loading experts on demand, delivering 20–24 tokens per second on a 48 GB M5 Pro at a 33 GB budget. It works offline, exposes OpenAI/Ollama/Anthropic-compatible APIs, and supports coding agents like Claude Code.

A minimal AI software engineering agent that uses only bash commands to solve coding tasks, providing a high-performance, hackable alternative to complex agent frameworks.

A plugin for DeepSeek Harness that turns a single agent session into a coordinated team of sub-agents capable of planning, delegating, and executing complex, multi-step tasks.

Eta is an Android app that turns a phone into a system‑level AI agent. Users supply their own LLM API key; the built‑in Agent Runtime lets the model call Android APIs, a GUI accessibility agent, a Linux‑style terminal, file tools, a WebView browser, and remote services. It can set alarms, control media, read notifications, run scripts, edit files, and more, with optional root/Xposed integration for deeper system access. Designed for power users and developers who want programmable, cross‑app automation on Android.

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.

An AI-powered short-drama production platform that automates the entire pipeline from script generation and character design to storyboard breakdown and final video export.

A CLI for git worktree management designed to let developers and AI agents run multiple parallel changes in separate directories without the friction of native git commands.

A curated library of reusable, procedural 'skills' and workflows designed to help AI coding agents build consistent web interfaces, 3D effects, and game systems.

A personal essay/guide on maximizing AI-assisted coding by acting as a manager — covering environment setup, prompt discipline, avoiding poor-quality AI output, and lifestyle integration. Written by an experienced user, with no code of its own.

A collection of reference implementations for the Model Context Protocol, designed to help developers learn how to build secure, standardized tool and data integrations for Large Language Models.

A privacy-focused, self-hosted AI answering engine that combines private web search via SearxNG with local or cloud LLMs to provide cited answers.

REA (Reverse Engineer Anything) is a Node.js platform that lets AI agents automatically analyze binaries, JavaScript/Electron apps, .NET assemblies, Android APKs, firmware, websites, and more. It runs traditional reverse‑engineering tools (Hopper, Ghidra, IDA, pwntools, JADX, Binwalk, browsers) behind a local MCP server, returns structured evidence (pseudocode, module graphs, network data), and enables the LLM to explain or re‑implement the observed behavior. Setup is done via `npx rea‑agents setup`, and the same capabilities are available through a CLI. The project targets lawful reverse‑engineering research and integrates with agents such as Claude, Gemini, and Grok.

An open-source LLM observability and evaluation platform for AI agent tracing, prompt management, and production monitoring.

OpenCreator is a local-first AI workspace that combines agentic workflows with visual tools for video, image, and text creation, allowing users to manage complex media production tasks through a unified interface.

huashu‑skills is a curated index of 57 Claude‑compatible *Agent Skills* (design, data, video, persona, etc.). 22 lightweight skills live in this repo; the rest are separate repos. Install with a one‑line `git clone` (or copy a sub‑folder), restart the agent, and the skill becomes usable. An included updater checks for newer versions. The collection provides ready‑made, plug‑in functionality for AI agents without writing code.

Spirula Studio is an all-in-one, cross-platform application for training 3D Gaussian Splatting models from photos and videos without requiring complex Python or PyTorch setups.

QuantSkills is an open-source community directory of reusable quantitative finance tools and AI agent components designed to make research, data, and strategy validation transparent and reproducible.

A high-performance AI gateway that unifies 23+ AI providers under a single OpenAI-compatible API with automatic failover and semantic caching.

A library of 277 specialized AI expert roles with defined workflows and identities, designed for integration into AI programming tools to simulate a professional multi-departmental team.

An open-source AI agent platform for Android that integrates LLMs with system automation, a Linux terminal, and web agents to enable complex task execution on mobile.

Pipecat is an open‑source Python framework for building real‑time voice‑first and multimodal AI agents. It offers plug‑and‑play support for dozens of speech‑to‑text, LLM, and text‑to‑speech services, composable pipelines, and built‑in multi‑agent orchestration. The ecosystem includes client SDKs (JS, React, Swift, Kotlin, C++, ESP32), a CLI for scaffolding projects, a UI component library, a debugger (Whisker), and many example bots. Ideal for developers and researchers who want to prototype or deploy voice‑centric conversational systems quickly.

Doop is an open‑source, real‑time collaborative design canvas where humans and AI agents (Claude, OpenAI, Anthropic, etc.) can edit live HTML frames together. It offers multiplayer cursors, page organization, AI‑driven design cards, optional integrations, and can be self‑hosted via Docker or bun.

A free 9-week data engineering course from DataTalks.Club: build an end-to-end pipeline using Docker, Terraform, BigQuery, dbt, Spark, Kafka, and more. Available as a structured live cohort (graded homework + certificate) or fully self-paced. Requires basic coding and SQL; no data engineering experience needed.

A persistent file-based planning system for AI coding agents that stores task state on disk to prevent memory loss during context resets, crashes, or long-running sessions.

A Next.js‑based editor skill for AI agents that helps developers create and export App Store / Google Play marketing screenshots with themes, connected canvas, and multi‑platform support.

A native visual design system for DeepSeek Harness that allows AI to generate and manually refine editable websites, presentations, and videos.