Code agents

Autonomous judge agents fix your app's debt, build what's missing, and verify every change through a browser or gate suite before it lands, then find the next thing and keep going until you say stop. Each fixer owns a file-disjoint slice of the tree, so no two can ever collide, and nothing unverified reaches git. A Claude Code plugin.

Tracks which agent wrote which lines when multiple AI agents share one Git checkout.

A Claude Code orchestration protocol prototype for approved plans, bounded agent roles, and verified parallel work.

OpenCode + Ollama multi-agent workflow with builder/reviewer agents, headless runs, skills, validation gates, local session evidence, and redaction tools.

Production multi-agent Claude Code skill — turns a voiceover MP3 into a full whiteboard-explainer YouTube episode. Brief-on-disk + SHA256 verify, gated 2-wave generation, vision review loops.

小妍是一款面向科研学习与论文工作的 AI 研究助手。面对纷繁复杂的论文与信息，她帮你厘清研究脉络：输入一个主题，多 Agent 协作拆解学习路径、筛选必读论文，并生成结构化综述。 PDF 拖入即可分析，图表识别、复现建议与引用溯源无缝衔接；知识图谱则将论文、笔记与记忆编织为可检索的证据网络，让每个结论都有据可查。欢迎各位Star & PR。

Risk-adaptive Codex workflow with policy-level caps, native quality gates, and optional Superpowers handoff.

Ready-to-use AI agents for AgentsKit — shadcn-style: copy the source into your project. registry.agentskit.io

Agentic Ops Harness | Superpowers Like AI Harness for Ops. Currently works with Hermes Agent

Local-first governance for AI coding agents: the external, files-only record that verifies what your agents did, so no agent is the sole witness to its own work. One agent per lane, approvals, gates, and a hash-chained spine anyone can check against its published JSON Schema. Nothing leaves your machine.

Run and serve local LLMs everywhere .NET runs - desktop and the browser - on the SpawnDev.ILGPU.ML inference engine (GGUF models, KV-cache decode, WebGPU dispatch-plan capture/replay).

Think through a software project idea, then build it with one click.

Open agentic runtime in TypeScript — run any coding agent in the background, check its work before it lands, and project tools/skills/agents/workflows into CLI, HTTP, MCP, SDK, Mastra & the AI SDK.

一套由本地 Markdown 驱动的长期 AI 协作模板，让背景、此刻、方向与选择不再随对话丢失。

Self-hosted AI life OS — private AI with persistent memory, tasks, calendar, household & team sharing, n8n automations. Your AI. Your data. Your server.

Simple OpenAI GPT API reference chatbot, that demonstraits basic turn based chatbot principles.

Image generation CLI that reuses your Codex (ChatGPT) login.

Перегоняет книгу, курс или подкаст в набор вызываемых Claude Code скиллов. Русская локализация cangjie-skill с тройной проверкой и приёмкой живой маршрутизации.

Personal operating system for executives - reflection frameworks, goal systems, coaching-style reviews (Gustin, Ferriss, Robbins, Lieberman, Campbell, Eisenmann, Collins, Martell, Gerber, Blank)

Approve Claude Code permission prompts right from your menu bar — no terminal switching — plus live status for every AI coding agent (Claude, Codex, Copilot, Antigravity).

A Claude skill that audits or builds any website/UI against ~60 research-backed conversion & design-psychology principles. Auto-detects live URL vs local code.

A tiny and fast coding agent written in Rust.

Unattended PRD delivery for Claude Code

A browser-based dashboard for measuring the impact of Claude Code optimization tools (Graphify, Claude-Mem, opusplan) using ccusage session data. No install required.

Retroactive Claude Code Skill usage analytics. Single-file Python CLI, reads your existing ~/.claude transcripts, counts real Skill tool_use invocations (not prose mentions). No hook to install ahead of time.

DARE is a structured methodology for AI-assisted software development. Four phases (Design → Architect → Review → Execute) keep humans in control while AI handles execution. Built for Cursor IDE, language-agnostic, battle-tested in production.

Keyless-first MCP server — 150 tools for US federal + state/local (SLED) government contracting, spending, regulation & partner vetting. SAM.gov, USAspending, Grants.gov, OFAC, FDIC, EPA, CourtListener + 45 more. No API key. Honesty-hardened.

macOS menu bar app for Claude Code token usage visualization

A native macOS blogging app you can talk to. Write in Markdown, build your site by describing it to Claude, edit drafts with AI, and publish free on GitHub Pages.

Real-time biofeedback engine for macOS — reads cognitive overwhelm from your face and webcam at 30 FPS, then lets Claude restructure your workspace.

The verification layer for autonomous agents — an independent, capital-aware verdict before an irreversible action (/review), a cryptographically signed proof after, and a public on-chain track record, losses included. Free EdgeProof backtest validation. Underneath: memory, sandboxed execution, marketplace. Pay in sats, USDC (x402), or card.

SEO-focused 12-skill curated subset of claude-skills. For SEO consultants and freelancers using Claude Code.

Ferrox - High-performance, stateless LLM API gateway written in Rust. Sits in front of LLM provider APIs (Anthropic, OpenAI, Google Gemini, AWS Bedrock) and exposes a single OpenAI-compatible HTTP API surface. Inspired by nginx/envoy proxy design — stateless, lock-free, horizontally scalable.

Local-first persistent memory for AI coding agents (Claude Code/MCP). Reliable semantic recall — $0, offline, open source.

Self-hostable email for humans and agents on Cloudflare: send, receive, search, MCP, IMAP, and webmail. Identity-bound credentials (v1.4).

The open benchmark for measuring the reliability, reproducibility, and determinism of AI context systems. Transparent metrics, reproducible tests, vendor-neutral results. Context Trust Levels CTL 0-4.

Research framework evaluating retrieval-augmented vision-language models for evidence-grounded medical visual question answering.

Seedance 2.5 Preview MCP server powered by MuAPI.

Tako MCP gives your agent industry-leading live web search plus licensed data that the open web does not have. That includes company financials, macroeconomic indicators, web and app traffic, sports, US government spending, and more. Hosted MCP server, free tier, no API key to start.

A deterministic, AST-verified call graph for any Python project — every edge traceable to a file:line. The reproducible, CI-gateable alternative to LLM-generated diagrams. No code executed. Nothing leaves your machine.

🐀 the agentic stack for those who have to save tokens

An open-source AI Software Engineer CLI and terminal coding assistant built in Go.

Finyuus is a code-first platform for building, running, and governing AI workflows. It introduces a small indentation-based DSL for composing agents, tools, guards, human approvals, and nested workflows — then runs them.

Artifact repository for the paper "ReCodeAgent: A Multi-agent Workflow for Language-Agnostic Translation and Validation of Large-Scale Repositories", In Proceedings of The 41st IEEE/ACM International Conference on Automated Software Engineering (ASE 2026), Munich, Germany, October 2026

Modality-routed retrieval for long-term multimodal agentic memory. Code for arXiv:2608.01543.

ACP-native multi-agent living group chat — event-driven, agent-agnostic, formally reviewed

Skainet Bridge — local automation bridge for AI agents: one process, one port. MCP server + client, REST, browser extension, sandboxed exec. Independent project, unaffiliated.

Your AI marketing team for B2B SaaS: 72 specialist agents and 19 Claude Code skills — SEO/AEO/GEO, content, paid media, newsletter sponsorships & content syndication, email, PLG, pricing, ABM accounts, B2B podcast, product marketing, sales enablement, analytics — run by the CATALYST orchestrator. MIT; works in Cursor, Copilot and Windsurf.

Cross-platform desktop app for Roku development: device discovery, terminal sessions, remote control, sideloading, screenshot capture, automation scripting, and a built-in AI assistant with Gemini, Claude, Codex, and Copilot

FastMCP server for Music Assistant — control MA from Claude Code, Codex, Cursor, and other AI agents

Agentic Spec-Driven Development: a contribution pipeline for software built with AI agents. Every change disclosed, gated by hard checks, reviewed against its spec, and merged by a named human.

Autonomous CI/CD for AI-written code: plans, builds, and adversarially audits changes to any repo — a different model reviews the builder — shipping only what passes deterministic gates. Failures become durable lessons; runs are checkpoint-resumable & context-optimized. Claude Code · Codex · Gemini.

Skills and more for open code free models

The open-source AI content factory that runs on your PC — discovers topics, researches, writes, reviews with multi-model QA, and publishes. Local-first via Ollama, $0 API costs.

Agent CLI, Claude Code plugin, and skills for Tenjin, the x402-native knowledge marketplace. USDC on Base, no API keys.

Obsidian vault template that gives AI agents persistent, shared memory — Claude Code, Codex, and Cursor read and write the same second brain. AGENTS.md rules, per-project memory, human-gated principles.

A live diff monitor for the terminal pane beside your coding agent. Built to stay open, not opened once per review.

Quira is a next-gen Python RAG framework featuring speculative vector search, Context Tetris for token compression, and differential caching to eliminate redundant API costs.

Turn messy tasks into controllable Agent Loops.

Model-agnostic, Claude-Code-native SDR orchestrator: research → enrich → outreach over your own data-provider keys — fully local, zero Google. A Claude Code plugin (orchestrator skill + phase sub-agents + bundled MCP server) and a standalone CLI.

MCP-native CVE, PoC & exploit discovery toolkit (Metasploit, Nuclei, GitHub & more)

Custom Claude Code plugin marketplace with skills for FARM stack, open source contributions, repo maintenance, and portfolio management

🐱 Knocks over your AI agent's "done" claims with hidden retests, mutation gates, and executable acceptance contracts for Java. If it's hollow, it falls.

Give Claude Code a persistent mind: an autonomous agent framework that sets its own goals, forms predictions, and learns from outcomes across sessions.

Multi-agent AI debate app where model teams argue, challenge claims, track evidence, and produce judged verdicts with analytics.

A Kimi-web-style browser console for the Devin CLI — sessions, streaming, modes, models, usage, terminals — over the Agent Client Protocol (ACP)

Source-free architecture maps and deterministic polyglot refactoring for AI coding agents

Agent skill for Claude Code, Codex and Cursor: repeatable screenshots of GUI desktop apps on an isolated Xvfb display, driven via python-Xlib (XTEST) and annotated with ImageMagick

TypeScript SDK for writing AI agent plugins once and exporting them to multiple agent platforms.

TokenTelemetry launcher tab inside Hermes Dashboard — local observability for Hermes Agent and 9 coding agents.

Control your AI coding agents from anywhere — monitor sessions, approve permissions, and orchestrate Claude Code, Gemini, and Codex remotely.

AI-Powered B2B Outreach Plugin 2026: Persistent Briefing & 40 Skills for Claude Code

CSINT Research side project for static and staging-safe security regression checks of n8n AI workflows.

mirabile.app is an AI-powered web app that converts your dream career into a complete career roadmap.

138-module multi-analyst market intelligence system -- macro, cross-asset, geopolitical, alternative data analysts with FastAPI, multi-AI providers, ML, and Telegram/email alerts

Turn vague agent prompts into precise, verifiable XML specs — then execute them. Headless multi-CLI generator skill (Claude, Grok, Codex, Gemini). Agent Skills + skills.sh compatible.

Install Stet CLI and agent skill — change control for AI coding behavior

Taskspace — collaborative task manager built with DeepSpace

Kahoot-style live quiz platform built on the DeepSpace SDK

給非技術用戶與 AI agent 使用者的中文 meta instruction：重點不是讓 AI 盲目服從，而是要求先核實、少亂動、可驗收、出事可追溯。

Spend model capability like a scarce budget: tiered routing, spec-based handoffs, and a living playbook that carry strategy across sessions. A Claude Code plugin plus a portable method for using context economically, in code and beyond.

DASM：面向研发团队的 AI 协同开发框架，集成 Vibe Coding 工作流、编码智能体、多人协作工作台与桌面客户端，支持私有化大模型接入，一站式落地 AI 驱动的团队代码研发、项目编排与迭代管理。

A prompt bar that edits the site it's on — ask for a change; a real Claude Code session edits the repo and pushes to main. Self-hosted, single-owner, MIT.

This Is an ai agent that works on local machine. It is inspired by Jarvis. But the main key point of this System is it dont need any kind of Api key from any kind of provider or It also dont need any high end GPU to run it can even run on a 2 GB DDR 3 RAM laptop whole locally without internet.

Living architecture control plane for the AI-dev era. Derive real architecture from code, keep living docs + ADRs, enforce intent in CI, and steer AI agents to conform. Local-first, MIT.

Agent-first secrets, zero .env, zero servers. Encrypted per-env dotenv files committed to git via sops+age. Go + Zig CLI.

📖 Auto-generate feature-based documentation structure (AGENTS.md + docs/) for any project.

Project orchestration and AI agentic workflow utility.

Isolation for Claude Code on native Windows with MSVC, no docker/WSL.

Zero-token long-term memory for Claude Code. Indexes your own session transcripts and retrieves evidence deterministically — entity-context graph + temporal hierarchy, no LLM call anywhere in the memory pipeline. It tells you when it does not know.

Codex-led Qwen and DeepSeek AI team router via MCP, with token-efficient worker scripts and safety gates.

Agent Skills - 46 skills for AI agents (Sisyphus, opencode): frontend-perfection (Lighthouse audit), secret-scanner, security-review, commit-lint, coverage-analyzer, api-contract-testing, seo-toolkit, sql-helper, docs-system, code-review, github-repo-hygiene. skill.json manifests, CI

A software factory where AI agents propose and deterministic code disposes. A Python ADW engine runs coding agents in bounded phases; a Next.js cockpit observes and controls them. Both halves meet at one SQLite trace - the single source of truth.

Presence and coordination for several Claude Code sessions working in one project at the same time: a roster, names, messaging, and findings that outlive the session.

Official Code for AISE-Bench: A Full-Cycle Curated Benchmark for Information Seeking on Academic Knowledge Graphs (KDD 2026)

High-autonomy config for opencode: ship features end-to-end with one command. 5 models, 5 families, zero interruptions.

Twitch Chat Bot (Twitch чат-бот)

Hermes image generation provider for OpenAI-compatible endpoints and Codex Auth

FinOps Pilot is an autonomous procurement and payment-recovery agent for engineering teams.

Safely resumes interrupted Codex CLI turns and selected retry/wait states from tmux.