agents

CLI and JSX interface controlling Figma over the Chrome DevTools Protocol with 100+ read and write commands, no plugin install required.

Collection of DESIGN.md files derived from major brand design systems, intended as context for coding agents.

Curated system-prompt collection framed as a builder's reference for people designing their own agents.

Configuration framework adding thirty slash commands, twenty specialist agents, seven behavioural modes and eight MCP integrations to Claude Code.

Real-time guardrails that detect and block secret leakage in agent output and tool calls.

Agent workflow that drives Claude Code or Codex to build Godot, Bevy and Babylon.js games, running the engine and iterating on screenshots.

Planning layer that lets several coding agents share specs, tasks and decision history inside one workspace before they write code.

TypeScript WebGPU library with importable WGSL modules that renders the same shader to a browser canvas, headless Node.js, images, video, and CI snapshot tests.

Open registry for Eve agents where community agents install with one shadcn command and are published by pull request.

Workspace of databases, kanban boards, pages and calendars that Claude, Cursor and other agents read and write over MCP.

Playground repository with runnable side-by-side implementations comparing modern open-source agent frameworks on the same example tasks.

Bridge that lets coding agents drive arbitrary desktop software beyond the terminal and browser.

Sprawling toolkit bundling agents, skills, commands, plugins, hooks, rules, templates and MCP configs in one repo.

Claude Code skill that writes and runs Playwright automation on demand instead of exposing a fixed MCP tool surface.

Very large curated collection of Model Context Protocol servers organised by capability and integration.

CLI that runs coding agents inside hardware-isolated virtual machines for safe autonomous operation.

Reference site documenting one hundred and ninety-plus patterns for building and operating coding agents.

Reference library of hook-based skill activation, modular skills, ten specialist agents and documentation patterns that survive context resets.

Collection of system prompts and full tool definitions from Augment, Claude Code, Cursor, Devin, Kiro and VSCode Agent.

Focused collection of Claude Code plugins bundling specialised agents, skills and slash commands for development workflows.

Catalogue of agentic workflow patterns documented with Mermaid diagrams and implementation notes.

Curated index of AI agent frameworks, platforms and protocols with categorisation by use case.

Walkthrough of the Ralph technique: running a coding agent in a repeating loop against a spec until every item is finished.

Self-hosted Go sandbox service for agent-written code that reports the isolation tier of every run, refuses runs below a floor and brokers API calls without exposing credentials.

Mac app (Windows preview) where you hit a hotkey, talk, click and draw over your screen or a plugged-in iPhone, then paste the keyframes and transcript into your coding agent.

Read-only TUI that has Claude Code, Codex, Gemini or OpenCode turn a git branch into a step-through tour of intent, risks, questions and verdicts.

Composer for full-stack TypeScript projects built for coding agents: pick each layer in a web studio or CLI and get typed, linted, tested scaffolding.

Turns Claude Code, Codex and Mistral Vibe sessions into a live animated team call with characters, local voices, shared screens, sub-agent trees and replay.

Animated 3D TV-headed robot heads for React whose LED faces change expression to show what your AI agent is doing right now.

Markdown convention plus linter for AI-written project docs, catching stale versions and fields duplicated across files before agents read them back as context.

Open-source desktop inbox where coding agents pause for your decision on diffs, outgoing emails, or drafts, then continue with your feedback.

Linux command runner with a time limit that guarantees no process an agent's command started survives, using cgroup v2 kill or a subreaper plus pidfds.

CLI that asks a repository what it does and returns relevant files plus verbatim source excerpts for coding agents via Jev.

Local-first desktop app that tracks Claude Code and Codex tokens, cache hits, models, and cost across machines via a self-hosted hub.

Open-beta Cloudflare CLI generated from the full API OpenAPI surface, with JSON-first output, intent search, and TypeScript config for agents.

Small cross-platform code editor written in assembly with Vim mode, terminal panel, Git diffs, and optional agent side sessions.

Forty-three film styles as reusable prompts plus short films directed, drawn and scored entirely in code by Claude Opus, ready for agents to remix.

Prebuilt and DIY agents that use Jev, Laya, or Cua-S1 System-1 decision models for browser, desktop, games, and routing.

Workflow DSL that turns repeatable agent work into inspectable steps with Jev decisions and optional full agent loops.

Local-first Rust CLI and MCP that catches intent conflicts between parallel coding agents before they write conflicting code.

MIT-licensed desktop IDE where humans and coding agents share a canvas, jump from diagrams to code, and review AST-aware semantic diffs.

Turns the Figma desktop app into a local MCP server so agents get the full Plugin API with no cloud quota.

Research workspace where chat, files, PDFs, and notebooks share one dock, and the agent can spawn custom panes when needed.

Human-curated marketplace for paid and free agent skills across coding, writing, design, research, and ops.

Local-first desktop inbox for background AI agents, with unread results, approval queues, and bring-your-own model providers.

Multi-agent orchestrator that turns one goal into an ordered graph of coding-agent tasks with per-task runners and verified markers.

Agent-first CLI framework with one grammar and canonical names across SaaS providers, loading command schemas only when asked.

React and Tailwind registry of design-inspired components with agent install prompts, llms.txt, and shadcn CLI source.

Walkable Three.js worlds generated as code by agent swarms — Union Square, Kyoto Higashiyama, Death Star trench — no game engine.

Procedurally generated React Native agent avatars with nine mood loops, seven silhouette cuts, Skia rendering and animated SVG export.

Free WebGL shader orbs for React agent states (idle, thinking, speaking), tunable in a playground and installable via shadcn registry.

Local Rust daemon that schedules agent loops (Claude, Codex, Hermes, Pi) with REST, MCP, and a web UI — no cloud queue required.

Figma plugin that exports a selection as a local ZIP with previews, node geometry, SVGs and an offline browser, no API token.

Compiler-driven icon set for agents, MCP and RAG, with six validated drawings per concept and an MCP server of real names.

Source-linked archive of documented incidents where AI coding agents and chatbots deleted data, leaked secrets or burned money, each with the lesson.

Gallery of 89 real design systems with tokens and downloadable design.md files, plus an agent-readiness index and free token, contrast and grid tools.

Agent skill packaging 18 production CSS transitions plus commands to audit a project for hardcoded durations and swap them for motion tokens.

MCP server giving agents six design-lookup tools: UI inspiration galleries, component code, shaders and motion, plus scraping a site's type and palette.

Two-skill agent library: shader-driven WebGL UI components with real fallbacks, and Tailwind container queries used in place of viewport breakpoints.

Agent-driven design canvas running beside Cursor, VS Code, Claude Code or Codex, storing designs in an open format that lives inside your repo.

MCP server giving agents persistent memory as Open Knowledge Format markdown files on disk, indexed in SQLite FTS5 for fast lookups.

Emacs shell for chatting with any ACP-driven agent — Gemini CLI, Claude Agent, Auggie, Mistral Vibe — inside a native buffer.

Native Mac terminal that classifies every session as needs-you, working, idle, finished or failed, and queues the ones awaiting a decision.

Library of DESIGN.md files capturing real brand design systems — colour, type, spacing — in a format coding agents can read directly.

Paid Sanity + Next.js 16 / Astro 7 boilerplate built for agents: AGENTS.md, scoped skills, two in-repo MCP servers, llms.txt, page builder, fetch layer.

Copy-paste components for agent interfaces: chat, streaming text, thinking states, human-in-the-loop approvals, task lists and result rendering.

Steering behaviour and spatial query library for 3D agents, usable with Three.js for flocking and pursuit.

Desktop platform where specialised frontend, backend and database agents collaborate to generate full applications from a plain description.