Dự án mã nguồn mở Developer Tools tốt nhất

Skills For Real Engineers packages Matt Pocock's own agent workflow into roughly 25 slash-command skills covering requirements-gathering, TDD, debugging, and codebase-architecture review. Install the managed Claude Code plugin to track his updates automatically, or pull editable files via skills.sh and fork them. It doesn't automate development for you — it disciplines the AI coding agent you already run.

ECC is an MIT-licensed system from affaan-m/ECC that installs into AI coding agent harnesses such as Claude Code, Codex, and several others, adding skills, specialized agents, persistent memory, and AgentShield security scanning around whatever large language model the harness already runs. The README positions it as developer productivity tooling for LLM agent development and agent performance tuning rather than a code generation tool itself, composing around whichever model the harness already calls.

trimstray/the-book-of-secret-knowledge is a highly-starred GitHub repository offering a curated collection of resources for system and network administrators, DevOps, pentesters, and security researchers, though it aims to be useful for everyone. It compiles inspiring lists, manuals, cheatsheets, blogs, hacks, one-liners, and CLI/web tools, organized into main chapters such as CLI Tools, GUI Tools, Systems/Services, Networks, Hacking/Penetration Testing, and Shell One-liners. The project emphasizes quality over quantity, welcoming contributions that are inviting, clear, not tiring, and useful, with an active RSS feed for updates.

yt-dlp is a Python command-line downloader that pulls audio and video from thousands of sites and hands post-processing off to ffmpeg. It's a fork of youtube-dl built for people who want fast fixes when a site breaks, not a polished app.

Ollama is a Go-based CLI and local server for running open large language models — Llama, Gemma, Qwen, DeepSeek, and more — on your own hardware. It wraps the llama.cpp inference engine, exposes a REST API on port 11434, and ships official Python and JavaScript clients so apps can call a local model instead of a hosted one.

GitHub gitignore templates is the collection maintained in the github/gitignore repository, licensed CC0-1.0. GitHub uses this exact list to fill the .gitignore chooser shown when you create a new repository or file, and it splits templates into a curated root set for common languages, a Global folder for editors and operating systems, and a community folder for narrower or version-specific cases.

Ponytail is a skill that injects a code-minimalism ruleset into AI coding agents, teaching them to apply a seven-rung decision ladder before every code change. In benchmark tests against Claude Code editing a real FastAPI + React repo, it cut 54% of lines on average (up to 94% where agents over-built) while keeping every safety guard intact. Supports 16+ AI coding platforms and requires no config.

Spec Kit (github/spec-kit) is GitHub's open source toolkit for Spec-Driven Development, distributed as the Specify CLI plus a set of AI-agent slash commands. Rather than generating code from a single prompt, it walks a project through a fixed sequence (constitution, specify, plan, tasks, implement), treating the written specification as the artifact that actually gets executed, not just a document written before the coding starts.

30 seconds of code (Chalarangelo/30-seconds-of-code) is a GitHub repository that points to a collection of coding articles at 30secondsofcode.org, maintained by Angelos Chalaris. Each article pairs a code snippet with an explanation and an example, and the snippets themselves are licensed CC-BY-4.0; the project isn't taking new content submissions right now, per its README.

Awesome DESIGN.md is VoltAgent's collection of 73 ready-made DESIGN.md files — plain-text design-system briefs extracted from real brand websites like Stripe, Tesla, and Notion — built so an AI coding agent can generate UI that matches a specific visual language instead of a generic default look.

Pi Agent Harness is a TypeScript monorepo that bundles everything you need to build and run AI agents: a unified LLM API supporting OpenAI, Anthropic, and Google; an agent runtime handling tool calling and state management; a self-extensible coding agent CLI; and a terminal UI with differential rendering. The README documents three containerization patterns for stronger isolation since no built-in permission system exists. Installation is via npm with supply chain hardening baked in — pinned deps, exact versions, and a shrinkwrap for the published CLI. It's MIT licensed and actively maintained.

Agent Skills is an open-source AI agent skills collection from Addy Osmani: 24 structured workflows that push AI coding agents through the spec-plan-build-test-review-ship discipline a senior engineer would follow. Instead of one-off prompts, each skill is a step-by-step process with verification gates, so the agent writes a PRD before code and proves tests pass before shipping.

Windows Terminal is Microsoft's open-source terminal application, developed in the microsoft/terminal GitHub repository (MIT license) alongside the original Windows Console Host. It adds tabs, rich text, globalization, configurable profiles, and theming on top of the console infrastructure the same repo maintains, and can be installed from the Microsoft Store, GitHub Releases, winget, or the unofficial Chocolatey and Scoop packages.

Neovim is an open-source project at neovim/neovim that refactors Vim's internals to make the editor easier to maintain, easier to split across contributors, and open to modern interfaces without patching the core. It ships a plugin API reachable from languages including Lua, Python, Ruby, Go, and JavaScript/Node.js, an embedded terminal emulator, and asynchronous job control, while staying compatible with most existing Vim plugins.

Claude-Mem is an open-source plugin at thedotmack/claude-mem that adds persistent memory to Claude Code and several other AI agents. Per its own description, it records what happens during a session, runs an AI compression step on those records, and surfaces the useful parts again once a new session starts, storing everything in SQLite with Chroma layered on for semantic lookup.

Every Programmer Should Know (mtdvio/every-programmer-should-know) is a GitHub collection of links to books, talks, papers, and articles the author considers relevant to programmers regardless of stack or seniority. It is licensed CC-BY-4.0, and groups its links into topic sections spanning CS fundamentals, distributed systems, security, architecture, career development, and soft skills, explicitly presented in no particular order.

OpenCut is an open-source video editor, presented as an alternative to CapCut, that is currently undergoing a complete rewrite. While the active, usable version remains at `opencut-app/opencut-classic` and `opencut.app`, this repository details the ongoing development of the new architecture. The planned rewrite will feature an Editor API, a plugin-first architecture for third-party plugins, and a unified Rust core to support web, desktop, and mobile platforms from a single codebase. Future additions include an MCP server for AI agents, a headless mode for automation and batch rendering, and an in-editor scripting tab. The project uses `proto` and `moon` for development, but external contributions are not yet being accepted as the architecture is still in design.

Svelte is an open-source compiler for building web applications, maintained at sveltejs/svelte under the MIT license with strong community support. Its README describes converting declarative components into JavaScript at build time, backed by volunteer contributors, a public roadmap, and Open Collective funding for costs like hosting.

Understand Anything is an open-source Claude Code plugin maintained by Egonex; the project traces back to an original build by Lum1104. It runs a multi-agent pipeline over a codebase and produces an interactive knowledge graph of files, functions, classes, and dependencies. The GitHub repo lists TypeScript as the primary language, and an MIT license. It also installs into Cursor, VS Code Copilot, Codex, Gemini CLI, and several other AI coding platforms via a one-line script.

DevOps Exercises is a GitHub repository collecting 2624 questions and exercises spanning Linux, networking, Kubernetes, Terraform, and cloud platforms, written as plain markdown Q&A with no app or course wrapped around it.

deterministic MP4 videos from HTML, CSS, media, and seekable animations

OmniRoute is a self-hostable AI gateway that puts one endpoint in front of many model providers. You point coding agents like Claude Code, Cursor, or Cline at it, and it routes requests across providers — including free tiers — with automatic fallback when one is unavailable.

Awesome Claude Code is a GitHub-hosted awesome list for Anthropic's Claude Code CLI: dozens of annotated entries split across categories from beginner onboarding to voice I/O and MCP infrastructure. Every pick gets a real description instead of a bare link, and the list is mid-rewrite, so recently-added and legacy sections coexist.

Claude-Mem is an open-source plugin at thedotmack/claude-mem that adds persistent memory to Claude Code and several other AI agents. Per its own description, it records what happens during a session, runs an AI compression step on those records, and surfaces the useful parts again once a new session starts, storing everything in SQLite with Chroma layered on for semantic lookup.

Ollama is a Go-based CLI and local server for running open large language models — Llama, Gemma, Qwen, DeepSeek, and more — on your own hardware. It wraps the llama.cpp inference engine, exposes a REST API on port 11434, and ships official Python and JavaScript clients so apps can call a local model instead of a hosted one.

DwarfStar is a specialized native inference engine optimized for specific DeepSeek V4 Flash, DeepSeek V4 PRO, and GLM 5.2 models. It targets local execution on consumer hardware, supporting Metal on Macs (including SSD streaming for lower-memory machines), NVIDIA CUDA (with multi-GPU and micro-batching for multi-user servers), and ROCm on Strix Halo systems. The engine is self-contained and deliberately narrow, focusing on a few models rather than being a general GGUF runner, with features like model loading, prompt rendering, tool calls, and an HTTP server built together. It also supports tensor parallelism across multiple Macs and pipeline parallelism to run larger models or accelerate prefill. The project acknowledges strong assistance from AI models in its development, while crediting `llama.cpp` and GGML for foundational engineering knowledge and components. Currently, the software is in beta, undergoing rapid changes, and is presented as a working template for users to adapt, potentially with coding agents, for their specific hardware setups.

witr is a Go-based CLI and TUI that traces a running process, open port, container, or held file back to the exact causal chain — supervisor, container, shell, or session — that put it there, in one command instead of five.

Pi Agent Harness is a TypeScript monorepo that bundles everything you need to build and run AI agents: a unified LLM API supporting OpenAI, Anthropic, and Google; an agent runtime handling tool calling and state management; a self-extensible coding agent CLI; and a terminal UI with differential rendering. The README documents three containerization patterns for stronger isolation since no built-in permission system exists. Installation is via npm with supply chain hardening baked in — pinned deps, exact versions, and a shrinkwrap for the published CLI. It's MIT licensed and actively maintained.

Node.js/Typescript/Bun Production Process Manager with a built-in Load Balancer.

A curated list of awesome command-line frameworks, toolkits, guides and gizmos. Inspired by awesome-php.

Glances an Eye on your system. A top/htop alternative for GNU/Linux, BSD, macOS and Windows operating systems.

A syntax-highlighting pager for git, diff, grep, rg --json, and blame output

This repository started out as a learning in public project for myself and has now become a structured learning map for many in the community. We have 3 years under our belt covering all things DevOps, including Principles, Processes, Tooling and Use Cases surrounding this vast topic.

🎨 A curated list of delightful VS Code packages and resources.

Our next-generation, interactive app is currently in beta! Come try it out at terminalthemes.com! Over 750 terminal color schemes/themes for iTerm, with ports to Terminal, Konsole, PuTTY, Xresources, XRDB, Remmina, Termite, XFCE, Tilda, FreeBSD VT, Terminator, Kitty, Windows, Visual Studio,, Ghostty, and dozens more

WinGet is the Windows Package Manager. This project includes a CLI (Command Line Interface), PowerShell modules, and a COM (Component Object Model) API (Application Programming Interface).

最新Docker容器技术，从真实案例中学习最佳实践！| Learn and understand Docker&Container technologies, with real DevOps practice!

Oh my tmux! My self-contained, pretty & versatile tmux configuration made with 💛🩷💙🖤❤️🤍

Nodejs extension host for vim & neovim, load extensions like VSCode and host language servers.

A community-driven Emacs distribution - The best editor is neither Emacs nor Vim, it's Emacs *and* Vim!