Software Engineer
Things I use, things I recommend, things that save time.
How to stop getting rate-limited after an hour. Context window mechanics, the /effort command, model routing, and the silent drains most people miss. Applies to Claude Code, Codex, and Copilot.
UI/UX Pro Max for styles, Impeccable for anti-patterns, Google Stitch for design tokens. Then build your own brand skill and a Storybook of components. That is the whole stack.
CLI that reads your git diff, generates a test plan via AI, and executes it in a real browser with Playwright. Extracts cookies from your local browser for authenticated sessions.
Local semantic search and graph RAG that runs on your filesystem. No vector database. No cloud service. Just fast, file-based knowledge retrieval that syncs however you want.
Google Labs shipped an AI design tool that generates high-fidelity UI from text prompts, exports real code, and uses DESIGN.md to enforce your design system across any AI coding tool.
A Go binary that gives AI agents browser control via REST API. Serves the Accessibility Tree instead of raw HTML, cutting tokens by up to 90%. No Node, no Python, no dependencies.
WebSocket bridge connecting AI agents to your already-running webapp. Zero browser startup, real-time DOM snapshots, console logs, and network visibility.
Rust TUI, 71 skills, 37 agents, and a knowledge system that deploys to Claude Code, Codex, Copilot, Gemini CLI, Cursor, and five more platforms. Context engineering as a portable engineering artifact.
The tools behind my security agent. Shannon for white-box pentesting, PentAGI for autonomous scanning, Ghost Security for code analysis, Argus and WebCopilot for recon. All wired into one agent in my wololo setup.
Builds a local knowledge graph of your codebase with Tree-sitter so AI assistants read only the files that matter. Claims 6.8x fewer tokens on reviews.
Rust-built agent operating system. 137K lines of code. Single 32MB binary. Seven autonomous Hands, 40 channel adapters, 38 tools, and 16 security systems. Not a wrapper.
Free static hosting with no signup, built for AI agents. Publish presentations, dashboards, and prototypes to a live URL in seconds via API.
notebooklm-py and the NotebookLM skill for Claude Code. Source-grounded, citation-backed answers from your own knowledge base, zero hallucinations.
Line-level AI-vs-human code attribution. Captures agent transcripts in commits, tracks which lines were written by Claude, Codex, or you. Founded by ex-GitHub CEO with $60M seed.
justspeaktoit, Parler, and Wispr Flow. Three takes on voice-driven AI coding, from free and brilliant to polished and paid.
Community-driven fork of OpenAI's Codex CLI with multi-agent orchestration, browser integration, theming, and Claude/Gemini support. What Codex should have been.
Markdown-based programming language for long-running AI sessions. Contract-based semantics, parallel execution, and composable workflows. The thesis: an AI session is a computer, and this is its language.
From DIY tmux+Tailscale to dedicated platforms like conductor.build and vibetunnel. Every approach to running coding agents remotely, rated.
Zig-based headless browser engine designed from scratch for AI agent automation. No GPU compositor, no extension runtime, just the bits you actually test against.
Chrome extension and gateway that relays browser control between AI agents and your actual browser session. No headless browsers, no separate instances.
Free open-source screen recorder with auto zoom, annotations, cursor tracking, and motion blur. Does what Screen Studio does for $89, but costs nothing.
Give your coding agent a phone number. SIP, ElevenLabs voice, call-back alerts. The STT-to-model-to-TTS loop is still too slow for real work, but the demo is unbeatable.
Claude Code skill that gives agents a sandboxed browser via QuickJS WASM. Persistent page state, full Playwright Page API, 30% faster and 40% cheaper than Playwright MCP.
Built by Obsidian CEO Steph Ango. 22K stars. Gives Claude Code full read-write access to your vault. Search, create, link, and organise notes through your coding agent.
Meta-prompting framework with 69 commands and 24 agents that solves context rot in long AI coding sessions. 48K stars and growing. Best starter pack going.
A skill that installs frontend design taste into your AI coding tool. 20 slash commands for typography, colour, layout, and motion. Teaches Claude Code to stop defaulting to Inter and purple gradients.
50+ UI styles, 97 colour palettes, 57 font pairings, and a design system generator. The anti-AI-slop skill with 60K+ stars.
TypeScript SDK adding act, extract, and agent primitives to Playwright. v3 is CDP-native, 44% faster, and the caching system replays actions at sub-100ms with zero LLM cost.
A curated markdown catalogue of 32+ AI writing tropes across 6 categories. Drop it into your system prompt and watch the "delve into the landscape" vanish overnight.
CLI that calls MCP tools directly without loading hundreds of tool definitions into your context window. The antidote to MCP context rot.
Meta-programming config framework for Claude Code. 30 slash commands, 20 agents, 7 behaviour modes. Turns vanilla Claude Code into an opinionated development platform.
Python scraping library with three fetcher tiers, adaptive selectors that survive site redesigns, and stealth that handles Cloudflare. 35K stars.
Package manager for the open agent skills ecosystem. Find, install, and share skills across Claude Code, Codex, and other coding agents with npx skills.
Push notifications for long-running agent tasks. ntfy for dead-simple HTTP pub/sub, PingMe for fanning out to Slack, Telegram, Discord, and basically everything else.
Terminal app that manages multiple Claude Code, Codex, and Aider instances simultaneously in isolated git worktrees. Parallel agents, zero merge conflicts. Written in Go.
Local-only meeting transcription using whisper.cpp and Ollama. Records, transcribes, summarises, and lets you query your meetings. Nothing hits the cloud.
Two tools for converting web pages to clean markdown. markdown.new runs on Cloudflare edge, Jina Reader uses a URL prefix. Both slash token usage by 80% or more.
Middleware proxy that routes Claude Code API requests to cheaper or alternative providers. DeepSeek, Gemini, Ollama, OpenRouter. Same brilliant UX, different models, lower bills.
Speech-to-text API with a generous free tier that turns voice into a viable input for coding agents. Fast enough you forget it is there.
The OG agent orchestration platform by rUv. SPARC methodology, 60+ agents, hive-mind swarms, all running on Claude Code. Before it became Ruflo and went full Rust.
CLI tool that parses your local Claude Code session files and shows you exactly where your tokens went. Daily breakdowns, per-model splits, 5-hour billing windows. Free and open source.
Microsoft tool that converts PDF, DOCX, PPTX, images, audio, HTML, and more into clean markdown. Perfect for feeding documents into your coding agent context.
Python framework that lets LLMs control browsers through a perceive-act loop. DOM-first approach, works with Claude, GPT, Gemini, and local models. 89% on WebVoyager.