acvnace/awesome-vibe-coding-resources：AI 编程智能体资源合集

• anthropics/claude-code - Coding agent that understands your codebase, automates tasks, explains code, and manages git, all via natural language.

• MyVibe - Instant deployment for vibe-coded apps via Claude Code.

• aider - AI pair programming in your terminal.

• codename goose - Local, on-machine AI Agent that allows you to use any LLM and add any MCP servers as extensions

• MyCoder.ai - Open source AI-powered coding assistant with Git and GitHub integration, featuring parallel execution and self-modification capabilities.

• codex-profiles - Small Bash helper for switching Codex CLI/Desktop accounts with isolated CODEX_HOME profiles.

• claude-dash - Real-time statusline for Claude Code showing context, cost, quota, cache, tools, and git status.

• agenttrace - Local TUI for inspecting AI coding-agent session logs, usage, cost, latency, tool failures, diffs, and CI gates.

• OrcaReplay - Records a coding-agent session below the harness and replays it offline byte-for-byte with the network off, or forks it from any checkpoint onto a different model.

• nika - Workflow engine for AI: capture repeatable vibe-coded tasks as .nika.yaml files, statically checked before execution (schema, permits, cost), with tamper-evident traces after.

• ax - Local telemetry for AI coding agents.

• MUSE - Pure-Markdown memory OS for AI pair programming. Cross-conversation memory, 48 skills, role-based governance. Works with Claude Code, Cursor, Windsurf, OpenClaw, Gemini CLI, Codex CLI. Zero dependencies, MIT licensed.

• sober-coding - Post-generation audit for vibe-coded repos. 27 static checks across 7 dimensions (dead code, god files, duplicate blocks, deep nesting, empty except, dependency utilization…) with per-check fix suggestions and a CI mode that fails PRs on regressions. Language-agnostic, no LLM required.

• SwarmClaw - Self-hosted multi-agent runtime that delegates to Claude Code, Codex, Gemini CLI, OpenCode, Copilot CLI, Cursor Agent, Goose, Qwen Code, and Droid. Org chart view, schedules, runtime skills, persistent memory, sub-agent spawning. MCP-native (server and client). Electron desktop app, CLI, and Docker. MIT.

• SwarmVault - Local-first RAG knowledge vault with bundled MCP server (npx -y @swarmvaultai/cli mcp). Compiles raw sources (notes, transcripts, exports, docs, code) into a durable markdown wiki with a knowledge graph and hybrid SQLite FTS plus embeddings. Cuts agent token usage by serving compact wiki summaries instead of full file reads. MIT.

• Tree Ring Memory - Framework-agnostic local-first memory lifecycle for AI coding agents with Rust CLI, SQLite/FTS recall, forgetting, audit, consolidation, DOX/Revolve adapters, and TUI.

• Hyperconsciousness - Developer-alpha Rust CLI and MCP knowledge store for agent notes and task handoffs, with encrypted append-only records and scoped, expiring grants.

• DevIntern - Picks up tickets from Jira, Linear, Trello, Asana, Azure DevOps, GitHub Issues, or markdown files and turns them into self-reviewed pull requests using the coding agent of your choice (Claude Code, Codex, Cursor, OpenCode, Gemini CLI, and others), on your machines with your own model keys. A feasibility gate flags vague tickets back to the tracker with questions; the companion devpm tool creates codebase-grounded tickets from rough prompts. FSL-licensed, free for interactive use.

• whatbroke - Diff your AI agent's behavior between two runs: tool calls, args, cost, latency, and outcome flips, with flake detection and CI exit codes. MIT, on npm as whatbroke-cli.

• d1v - CLI deployment workflow for AI-built web projects with verified previews and production releases requiring explicit confirmation. MIT.

• MulmoTerminal - Browser grid of live Claude Code / Codex sessions started with one npx command. Each cell is a real PTY, colour-coded working / needs-you / done from the agent CLI's own hooks, with tmux-backed persistence and a git worktree per cell.

• YYLO - Command-line orchestrator for coding agents, repeatable workflows, and receipt-backed repository changes: typed task, validation, merge, and release-readiness boundaries, with each task on a dedicated branch/worktree and a merge queue that owns risk-based review. MIT, on npm as @yylo/cli.

• since-cutoff - Finds which APIs of your pinned Python dependencies changed after your coding agent's training cutoff, measures which of them the model actually gets wrong by type-checking its code against both versions, and writes one-line AGENTS.md / CLAUDE.md notes that are kept only if their example type-checks. uvx since-cutoff scan needs no API key. MIT, on PyPI as since-cutoff.

• agent-manager - Terminal UI that runs your installed Claude Code, Codex, OpenCode, Gemini CLI, and other coding-agent CLIs side by side, each in its own persistent tmux session. One list shows every session's live status, a session can spawn into its own Git worktree, and a full-file diff review sends line comments back to the agent. Apache-2.0, in homebrew-core as agent-manager.

• DropTheHassle - Puts an AI-built static site online from the terminal with npx -y dropthehassle deploy: a free HTTPS link without signing up (claim it to keep it), or your own domain. Also an MCP server for Claude Code, Cursor and Windsurf; the agent can never spend money. MIT.

• Ordewell - Command line planner and TUI that turns one goal into an ordered, editable plan of coding agent tasks, each with its own runner, model and mode, and marks a task done only when its completion marker appears in that runner's output. Apache-2.0, on npm as ordewell.

• Orbi - Self-hosted coding agent that turns a GitHub issue labelled ai-ready into an independently reviewed, merged PR and a tagged release; also works ops tickets. AGPL-3.0.

• mu - Coding agent built on Pi, with a CLI and a desktop app, where a small, fast judge model makes the routine calls at more than 30 decision points: which chunks of a long tool output enter the context, whether a rule-flagged command was asked for, whether a fetched page or MCP result carries instructions aimed at the model. Every verdict goes to a local ledger. MIT (the desktop app is Apache-2.0), on npm as mu-agent.

• Shipvela - Publish prebuilt static websites from the CLI or a connected coding agent, with GitHub updates, custom domains and HTTPS; a free Hobby plan is available.

• assay - Deterministic QA for web pages. It serves a page locally, drives every control it finds in Chromium through Playwright, and reports where the page contradicts itself, with no tests to write and no LLM. MIT, on PyPI as assay-ui.

AI-generated apps often ship with exposed secrets, open databases, or missing security headers. These tools help you catch issues before going live.

• AI Game Dev Framework - MIT-licensed documentation and agent workflow starter for game development, with design decision boundaries, project onboarding, and versioned delivery reviews.

• AI Coding Tools Guide - Workflow-based comparisons for Cursor alternatives, Claude Code, Windsurf, Cline, GitHub Copilot, Aider, Continue, and MCP-enabled AI coding.

• AI Book: AI Coding - Chinese AI coding handbook covering coding-agent workflows, tool selection, prompts, and practical AI-assisted development guides.

• CodeGuide - Creates detailed Documentation for your AI Coding Projects.

• LegacyDoc AI - VS Code extension that generates documentation, architecture maps, and AI-ready context packs for vibe-coded codebases.

• ContextStream - Shared project context for Cursor, Claude Code, Codex, Grok. Remote MCP: https://mcp.contextstream.io/mcp. Intelligence isn’t the bottleneck. Context is.

• Codex First Task Prompt Generator - Free browser tool for turning a vague Codex CLI project goal into a scoped first-task prompt with file boundaries, constraints, and acceptance checks.

• vibeprompt - Free 10-step vibe coding workflow with 56 prompts, 17 deep-dive articles, 46 field-tested fixes, and an interactive AGENTS.md + PRD generator. Open source, MIT licensed.

• Wikivibe - Practical knowledge base for AI-assisted development with guides, a glossary, jobs, and a public MCP endpoint.

• Vibe Coding with Confidence - Web handbook on building AI-assisted apps that work beyond the demo: plan, build, debug, harden, ship, operate, and scale.

• OrcaPromptVault - Dated archive of what the coding agents themselves are told: system prompts and tool schemas from 43 products, 44 of them recorded off the wire with a command that reproduces each. Useful when an AGENTS.md or CLAUDE.md is not having the effect you expect and you want to read what it is competing with.

• breaking-coding-chaos - Human-in-the-loop dual-loop control-plane skill suite for coding agents (Claude Code, Codex, Cursor, and more): throughline progress on disk, plan-spar alignment, then minimal clean-cut implement.

• Ariadne Loop - Local-first Loop Engineering workbench for writing verifiable AI coding-agent specs, verifier gates, rollback rules, and JSON reports.

• Cage - TypeScript contract harness that links Markdown specs to implementations, invariant-linked tests, and review freshness for Claude Code and Codex.

• Vibe Coding Essentials - Anti-hallucination guardrails for 9 frameworks + a 4-mode workflow system (Economy / Builder / Maintainer / Architect) for AI-assisted development.

• Fix AI UI Slop With Real UI Context - Free MIT anti-ui-slop Skill: write a product-specific design contract, cover required UI states, and run a hard finish gate; full UIZZE adds live reference search, validation, and audits across 800,000+ real web and iOS screens.

• BuildMuse - Curated interface videos, app screens, websites, motion references, and frontend effects for giving vibe-coding agents clearer visual direction.

• THROUGHLINE - Spec-driven development framework for coding agents, made of markdown and git with nothing to install. Personas debate each spec before code, and an append-only single source of truth keeps earlier decisions from drifting across sessions. Works with Claude Code, Codex, and Cursor. English and Korean, MIT.

• anti-slop-website-prompts - 15 free art-directed website build-prompts plus an anti-slop design skill for Lovable, Bolt, v0, Cursor, and Claude Code.

• Undercroft - Open-source CLAUDE.md operations contract that turns Claude Code into the maintainer of a plain-Markdown personal wiki: three-layer raw/wiki/schema architecture, checklist-driven ingestion, append-only changelog, and a worked example vault. Obsidian-friendly and portable to other agent CLIs via an AGENTS.md rename. MIT.

• Vibe Coding UI Specification - Free, MIT-licensed interface contract for AI-assisted React and Tailwind work, covering design tokens, component states, responsive behavior, WCAG 2.2 AA, reduced motion, production gates, and reusable YAML.

• Vibe-Coding Prompt Template - MIT-licensed five-step workflow (deep research, PRD, tech design, AGENTS.md, build) with copy-paste prompts and an npx vibeworkflow CLI that interviews you and writes the planning docs and agent files.

• liyupi/ai-guide - Free Chinese vibe coding handbook covering tool selection, prompting and context management, dozens of end-to-end project builds, and shipping to production. English and Traditional Chinese translations included. Online version.

• shadcn-vite-starter - MIT-licensed Vite and React starter whose CLAUDE.md, AGENTS.md and agent skills keep coding agents on its patterns: one folder per feature, shared data-fetching rules, tests first and a full check run before they finish.

• Bestax - React components for Bulma v1 that ship Agent Skills, an offline MCP server (bestax-mcp) and llms.txt docs.