Cursor 21

Skilloop is an Agent Skills directory for Claude Code, Codex & Cursor. Preview skill outputs before install. Chinese: https://skilloop.dev/zh | Skills: https://skilloop.dev/skills | Examples: /zh/ai-video/fable-5-5, /zh/ai-video/opus-5-5 | GitHub: https://github.com/xianyu110/skillhub

Skillry is a growing library of Agent Skills for AI coding agents. Every Skill carries a specific visual point of view, and you can see exactly what it makes before you install it — so what your agent ships is something you'd actually show. Web pages, slide decks, images and video. Install from the web or your CLI and use it inside Claude Code, Codex, Cursor and more. The taste is ours. The work is your agent's. Your agent can build a site, a deck, a video, an image. It just has no taste. Skillry is a growing library of Agent Skills that do. Each Skill carries a specific visual point of view, and you can see exactly what it makes before you install it — you are not buying a description. Install from the web, or authorize your browser once and run a single command in your CLI. Then invoke a Skill by name inside your agent and get a finished piece of work: a landing page that looks art-directed, a deck you would actually present, an image that stops the scroll, a short film that moves. Four kinds of work today — web pages, slide decks, images and video — with new Skills added every week. No API key required. It is not a marketplace. There is no community upload and no race to ten thousand listings. The library is deliberately small and grows one Skill at a time, which is the whole point: everything in it has a look worth keeping. Works with Claude Code, Codex, Cursor and more. The taste is ours. The work is your agent's.

AI-driven video editor for humans, developers, and coding agents Diffusion Studio is an open-source, AI-driven professional video editor for humans and coding agents. It combines an infinite canvas, timeline sequencing, media import, video editing, effects, audio tools, transcription, AI generation, and export in one workspace. Users can work in a browser, on macOS, or through the terminal, while agents such as Claude Code, Codex, and Cursor can analyze, edit, generate, and transcribe footage. The editor supports 4K playback and export, no-watermark exports, commercial use, and responsive canvas performance.

SaaS boilerplate with NextJS, Supabase, and Stripe integration for rapid SaaS development. SupaLaunch is a SaaS boilerplate based on NextJS and Supabase. It supports Stripe payments (one-time and subscriptions), database, authentication, storage, OpenAI APIs connection, design elements, landing page template, and email APIs. It's designed for vibe-coding SaaS applications with Cursor integration.

Shared project workspace for teams using AI coding agents Joxo is a shared collaboration project for teams using AI coding agents such as Claude Code, Codex, Cursor, Gemini CLI, and OpenCode. It connects each teammate’s existing coding-agent sessions through shared project memory, tasks, decisions, blockers, and handoffs. Joxo does not provide or resell AI models, subscriptions, or API access; each person uses their own AI provider account and computer. The service relays intentionally published project information between agents, helping teammates continue work without repeating context or sharing screenshots.

Open-source workspace for running and managing AI coding agents ADE is a local-first, open-source Agentic Development Environment that brings AI coding agents such as Claude Code, Codex, Cursor, Factory Droid, and OpenCode into one synchronized workspace. It runs across macOS, Windows, Linux, the web, terminal, and iOS, allowing developers to manage parallel coding tasks using isolated Git worktrees, agent chats, terminals, pull requests, built-in tools, and remote machines.

Codebase intelligence for documentation, AI context, health analysis, and safer software changes Repowise is codebase intelligence software that analyzes source code, dependencies, call relationships, Git history, ownership, tests, security signals, and architectural decisions. It converts this evidence into a reusable codebase model, source-linked documentation, dependency graphs, health scores, change-risk analysis, and grounded answers for developers and AI coding agents. Repowise supports MCP tools for Claude Code, Cursor, and other compatible clients, as well as hosted repositories, GitHub pull request reviews, team workspaces, and self-hosting through its AGPL-3.0 open-source core.

Collaborate with local AI agents in shared team rooms Alook is an AI agent collaboration platform that lets people and locally running AI coding agents share the same servers, channels, and direct messages. It gives agents persistent identities, handles, inboxes, memberships, and memory while keeping their runtime, tools, credentials, and codebase access on the user's own machine. Alook connects existing agents such as Claude Code, Codex, Cursor, OpenCode, and Pi to shared rooms so teams can communicate with and delegate work to them directly. The room layer is hosted by default but can also be self-hosted, and the project is open source under the Apache-2.0 license.

A community and resource hub for Cursor enthusiasts to enhance their experience. Cursor Directory is a platform for Cursor enthusiasts to explore and generate rules, browse MCP servers, post and follow the latest news, learn, connect, and discover jobs. It enhances the Cursor experience with custom rules and provides a community for users.

Monitor and control AI coding agents from your Mac notch Crew Tower is a native macOS control center for monitoring and managing AI coding agents from the MacBook notch or a floating pill on external displays. It supports agents such as Claude Code, Codex, Cursor, Gemini CLI, Qwen Code, and OpenCode. Users can review permission requests, approve or deny commands, answer agent questions, review plans, monitor progress and context usage, receive completion notifications, and jump directly to the correct terminal tab without leaving their current work. Crew Tower communicates locally through agent hooks and Unix sockets, keeping code, prompts, and transcripts on the Mac.

Unified API platform providing 700+ models and skills for AI agents. SkillBoss is a comprehensive API platform designed specifically for AI agents, developers, and product teams. it provides a single unified API access point to over 700 AI models and 1000+ specialized skills, including web scraping, search, voice generation, video creation, payments, and deployment. By consolidating services from over 61 vendors into one account, SkillBoss eliminates the need for managing multiple API keys and subscriptions. It is built to integrate seamlessly with popular AI coding agents like Claude Code, Cursor, Windsurf, and OpenClaw, offering a pay-as-you-go model where credits never expire.

Open-source workspace for managing multiple coding agents T3 Code is an open-source control plane for coding agents. It provides one interface for orchestrating Claude Code, Codex, OpenCode, Cursor, Grok, and other coding harnesses using subscriptions and credentials that users already own. The application supports agent workflows, model switching, GitHub branch and pull request management, inline diff review, self-hosting, forking, and custom tooling. It is available for Windows, Linux, iOS, and Android, and is MIT licensed.

Universal open-source package manager to deploy and manage skills across 32+ AI coding agents. SkillKit is a universal skill platform and open-source package manager specifically designed for AI coding agents. It serves as a bridge that allows developers to write agent instructions and capabilities once and deploy them across 32 different AI agents, including Claude, Cursor, Windsurf, and GitHub Copilot. The platform features 'Primer' for auto-generating instructions from a codebase, 'Memory' for persisting AI learnings across sessions, and a 'Mesh' network for inter-agent communication. With a library of over 15,000 skills and a powerful CLI, SkillKit simplifies the management, translation, and distribution of AI agent behaviors for both individual developers and enterprise teams.

Local CLI that reduces coding-agent token waste JFrog Boost is a free, local-first command-line tool that compresses noisy terminal output, MCP responses, file contents, and other tool data before it reaches coding agents such as Cursor, Claude Code, Codex, GitHub Copilot, and OpenCode. It reduces token usage while preserving important information such as errors, failures, summaries, and relevant context. Boost uses reversible compression and retrieval, allowing agents or developers to restore the original output when needed. It also provides graph-based code retrieval, file optimization, observability, telemetry, and enterprise-focused privacy.

AI token optimization stack for reducing agent costs and context Caveman is an AI efficiency operating stack that reduces the cost and context size of agent-native development. It provides an MIT-licensed skill, a free local proxy, an Agent SDK, and planned cloud and enterprise services. Caveman compresses logs, JSON, code, diffs, tables, tool outputs, and files before provider calls while preserving recoverable original bytes. It also supports caching, eval-gated model routing, token usage visibility, savings estimates, and verification workflows. The local wrap works with users' own provider keys and does not require a Caveman account.

Local desktop tool for monitoring and improving coding agents Blume is a desktop sidecar for monitoring and improving coding agent sessions locally. It works with Claude Code, Codex, Cursor, and other supported coding harnesses to show agent activity, track hidden rules and configuration files, monitor usage, and identify repeated corrections or workflows. Blume suggests updates such as new rules, skills, and verification steps, allowing users to preview, approve, dismiss, or save proposed changes.

AI website auditor that finds leaks and writes fix prompts Huncher is an AI website audit tool that scans your live site for conversion, design, SEO, and trust issues, then generates copy-paste prompts to fix them.

Native macOS server for fast local AI model inference oMLX is a native macOS inference server for running local AI models on Apple Silicon using Apple's MLX framework. It provides paged SSD KV caching to reduce time-to-first-token for coding agents, continuous batching for higher throughput, and OpenAI- and Anthropic-compatible APIs. oMLX supports LLM, vision-language, embedding, and reranker models, along with tool calling, MCP integration, multi-model serving, and a native menu bar application with a web dashboard.

Online platform for learning AI skills through courses and resources. Takeoff AI is an online platform dedicated to teaching AI skills. It offers courses and resources to help users learn how to build and work with AI, aiming to onboard the world to AI. The platform features courses, prompts, and a changelog, with options for both individual and team learning. It emphasizes practical application, particularly in coding with AI assistance using tools like Cursor.

A shared company brain for AI knowledge, skills, and workflows HQ is a shared company knowledge and workflow platform that provides the context layer for AI tools. It brings team knowledge, projects, reusable skills, workflows, secrets, and connected systems together in a workspace the company owns. HQ works across tools including Claude Code, Codex, Cursor, ChatGPT, and Claude, helping teammates and AI bots reuse shared context instead of starting from scratch.

ML.ai is an AI coding agent for VS Code, Cursor, desktop and your pull requests. It routes each step to the most cost-efficient model and shows every diff first. ML.ai is an AI coding agent that goes where your code goes. Use it inside VS Code and Cursor, in its own desktop app with no IDE required, from the terminal, or as a PR Reviewer that checks every pull request. Every surface runs on the same engine, with the same settings and the same memory of your repo. Most coding agents send every step to the same expensive frontier model. ML.ai hands each step to a focused agent and picks the most cost-efficient model that meets your quality bar, which typically cuts model spend by 30 to 45% on the same work. Explore reads and explains the codebase. Architect and Plan map the files, order of work and trade-offs. Only General can change code, and every edit or command waits for you to Allow or Deny it. The PR Reviewer summarizes each pull request, leaves line-level comments, and runs builds and tests before it recommends approve or request changes. It never merges on its own. Also included: reusable skills, background runs, MCP server support, a command safety classifier, and an access token that stays on your machine. ML.ai is a product of Pixis.ai, backed by SoftBank Vision Fund and General Atlantic.