Custom agents

Multi-Agent Code Evolution Platform: 4 LLMs compete, evolve, and breed code through a JIT compiler judge

内嵌进 MATLAB/Simulink 界面的 AI 助手侧边栏，支持 Claude Code/Codex 后端。

Infinite spatial canvas for terminal windows running AI coding agents. Wire agents together, pipe prompts, capture output to notes. Tauri + React.

本地优先的中文语音 AI 管家——说一句话，它替你操作 Windows（GLM-5.2 Agent + 本地 Whisper + GUI 自动化）

Agent-agnostic workflow skill for maintaining Obsidian knowledge bases

潜伏在终端与桌面里的幽灵 AI 编程 agent「Wraith」· A ghost-themed AI coding agent for terminal & desktop

Guardrails para desenvolver com agentes de IA: gate de CI bloqueante, hooks determinísticos, supply chain travada e fluxo de task com evidência. De 0 a funcionando sem orçamento

Agent skills for search, RAG, and document parsing with Mixedbread. Install with: npx skills add mixedbread-ai/skills

Agent interoperability and security, out of the box. The network layer underneath A2A: reachable, end-to-end sealed and payable agents without a server or public IP. MCP for coding agents, a2a-x402 payments, sandboxed execution.

An open source AI roleplay agent harness. One click builds an autonomous character agent that creates with you. Websites, code, music, videos, drawings, stories. Runs any SillyTavern card, deploys in one command, works in a phone browser.

a DSL for automated deterministic rules. Built for AI agents and quantitative Trading systems.

Not a chatbot. Not a coding assistant. A digital operator that controls your entire PC — 219 tools, screen perception, HID control, browser automation, Office suite, AI media. One sentence in, finished work out.

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.

Heddle — The policy-and-trust layer for MCP tool servers. Turn YAML configs into validated, policy-enforced MCP tools.

基于ChatGPT模型开发的AI工具微信小程序，提供聊天机器人、绘画助手等功能，支持用户通过文本和语音与 ChatGPT 交流，并且还具备画图功能，支持预览绘制的图片并可长按发送给微信好友。 WeChat Mini Program, an AI tool developed based on the ChatGPT model, provides functions such as chatbot and drawing assistant. It supports users to communicate with ChatGPT through text and voice, and also has drawing function.

Shepherd is an end-to-end AI Software Engineering platform that supervises the complete software delivery lifecycle—not just code generation, but planning, execution, testing, regression, governance, deployment, and continuous knowledge growth.｜Shepherd 将 AI 的一次性交付，转化为组织的长期软件工程资产。

Fully autonomous web crawler for security testing — maps endpoints, parameters, and security issues across traditional and SPA web applications. Drop a URL. Walk away.

Local-first ChatGPT MCP runtime for secure computer use, code, browser automation, and developer workflows.

将 AI 聊天接入通讯软件的连接与管理程序

مكتب عمل شخصي على الماك: يشتغل على مجلد تختاره أنت، وينفّذ مهامك المتكررة بمهارات تكتبها بنفسك، عبر النموذج الذي تربطه · A personal work desk for macOS: it works inside a folder you choose and runs your repeat tasks through skills you write yourself. Arabic & English. A صدقة — free and open source.

A brain for your mail: MailFathom turns IMAP mailboxes into a self-hosted, AI-native service. Mail synchronizes into your own PostgreSQL, is indexed for search and retrieval, and is served to AI agents over the Model Context Protocol. Semantic retrieval, answering, and gated write tools. .NET 10, AGPL-3.0-only.

Production-ready AI Business OS — 46 autonomous AI Employees, Graph RAG (Neo4j), multi-channel, BYOK across 17 LLM providers plus custom/local endpoints (Ollama), automatic fallback. 8 libs (Python+Java). Self-hosted, MIT licensed, guardrailed autonomy for regulated industries.

Uncensored-LLM is a fully air-gapped, zero-dependency, plug-and-play Local AI environment designed to run seamlessly from your local hard drive or a portable USB/SSD.

The only MCP server with live whale trades across 15 exchanges plus on-chain DEX flow. Smart Money Radar, real liquidations and outcome-scored signals. No API key required.

Open-source AI video platform for education. Type a topic, and SARAS writes the script, narrates it, generates the visuals, and hands back a finished, captioned lesson video — in your language, in minutes.

长任务，不刷屏：关键进度清晰可见，完成后自动折叠，详情随时展开。 Long agent tasks, without transcript clutter: focused progress, auto-folded history, details on demand.

BrokenMCP is a vulnerable implementation of the Model Context Protocol (MCP) for educational purposes.

TinyFish MCP server — web search, page fetching & extraction, browser automation and page monitoring for Claude, Cursor, VS Code and any MCP client. Search and Fetch are free.

On-device AI assistant for Termux/Android that watches Gmail over IMAP, drafts and sends replies with a local llama.cpp model, and is controlled entirely by text or email — no cloud AI, no API keys, no server.

A curated list of plugins for DeepSeek Harness (dsh) - DeepSeek Harness plugin ecosystem

Native WeShop Cordis plugin for DeepSeek Harness. Allow you to use infinite canvas with infinite creative skills.

交互式教程：20 节课，从「模型只会说话」讲到写出能装进真实 DeepSeek Harness 的插件。含可跑 demo 与在线版。

DSH (DeepSeek Harness) 专家模式 agent preset — 首席协调官 + 17位领域专家子代理 Expert-mode preset for DeepSeek Harness

The Zero-Trust Runtime Security & Governance Layer for Autonomous AI Agents.

Obsidian plugin that runs a local MCP server, exposing vault semantic search and file operations to AI assistants.

Chat Application using LLMs and Whisper for Speech-To-Text. Integrating Sherpa-ONNX for Text-to-Speech Support.

Screenshot-friendly Social Glass theme and workflow plugin for Herdr on macOS.

Official implementation of "FLARE: Verifying MILP Reformulations with LLM-Based Theorem Proving"

A structured agent runtime for LLMs. Structured context, multi-stage workflows, and ECS-based orchestration in a single binary

Lightweight CLI AI agent — chat, write code, and manage server software with natural language.

Local-first control plane for long-running and multi-agent AI systems. Observe, steer, recover and orchestrate agents.

Production-grade, security-first AI platform on a single local K8s cluster — GitOps (ArgoCD), Vault, Kyverno, Falco, self-hosted GPU LLM serving, and full observability. No cloud.

MCP server for flight price search via the Aviasales / Travelpayouts Data API. 13 read-only tools for Claude Code, Claude Desktop, Cursor and any MCP client: fares by route and month, cheapest day to fly, flexible dates, budget and inspiration search, plus airport/city/airline lookup that resolves place names to IATA codes.

让 MinerU 在 AMD 显卡上以 vLLM + hybrid-auto-engine 满血运行！ 本项目是对官方文档中 AMD/ROCm 生态支持的有效补充。针对 MinerU 3.x 提供了一套完整的 ROCm 7.x + PyTorch 2.11.0 + vLLM 源码编译与部署方案，解析质量与速度对标 NVIDIA 旗舰显卡。

Statewave use case in TypeScript: a grounded product advisor and support assistant, with an Ops Assistant that tracks and closes content gaps.

Bring Zendesk deep into your AI assistant: find answers and draft, update and translate Help Center articles, and manage Support tickets end to end: comments, triage and attachments. All in plain language.

Answers you can judge and act on without reading code — Claude Code plugins that keep implementation detail below a marked line. Off by default, armed one session at a time.

Evidence-driven multi-agent FPGA/SoC FPGA workflow for Codex: RTL, CDC/RDC, STA, verification, single-writer implementation, and independent sign-off.

AI-powered engineering IDE for Siemens TIA Portal and CODESYS: SCL code generation with Claude, GPT, Gemini or local Ollama, unit tests on PLCSIM Advanced, OPC UA, Git and MCP tools. TIA Portal V15-V21.

Bring Xnurta's advertising capabilities into your own AI assistant via MCP — natural-language Amazon ads data queries, analysis, and reporting.

Token-efficient facade and Codex skill for MCP for Unity: 96% smaller measured tool-schema surface, compact results, batching, and recovery.

DSH plugin: host-resident scheduled-jobs × AI-agent engine (hermes-agent cron inspired) — 60s ticker fires real agent sessions via cron; pinned-session / project-workdir / default-workspace targeting; timer_agent model tool + web board

Franklin is a LLM powered AI IRC chat bot

Open communication and coordination for AI agents.

Minimal harness-agnostic recursive language model layer — one binary, Python + llm()

🌐 DeepSeek 官方与开源生态插件市场 (deepseek.stream) 使用指南与一键安装协议规范

RAG for visually rich documents — layout, formula and table structure survive the pipeline. Ported from CUDA to Huawei Ascend NPUs and tested on Ascend 910C.

Practical engineering knowledge for building and operating production AI systems — LLMs, RAG, agents, inference, evaluation, system design, and AI infrastructure.

Zentra is an open source CLI agent harness tool designed to assist with SAST and DAST on your application

Open-source, supervised AI browser agent for Chrome. Plan, navigate, click, type, and verify web tasks from a side panel with your own OpenRouter key.

Agentic AI monitoring and observability — OTEL traces and local log files from Copilot, Claude Code, Codex, and OpenCode. Token usage, cost, summaries, and efficiency insights. VS Code extension, npx, and Docker.

Local-first coding agent with operator-owned prompts, memory, plans, tools, and multi-agent workflows in the terminal.

Browser automation for AI agents in your real, logged-in Chrome. You watch every action and can veto any tab. Chrome extension + zero-dependency CLI: works with any agent, no MCP server, no relaunch.

See what your agent sees. Self-hosted Chrome for AI agents with bounded API control, live noVNC takeover, npm/PyPI clients, and MCP servers.

The Hoppscotch MCP Server implements the Model Context Protocol (MCP) to connect AI agents and coding assistants including Claude Code, Cursor, VS Code Copilot, GitHub Copilot CLI, and Gemini CLI directly to your Hoppscotch workspaces, collections, specifications, and environments.

The Fable Method: a structured problem-solving loop that gives any model Fable 5 grade thinking. Four Hermes Agent skills (think/act/prove/grow) + adversarial eval.

Open-source, self-hosted AI agent for terminal, Telegram and Raspberry Pi. Local Ollama or Claude/Codex workers, enforced by a deterministic capability kernel, scoped approvals and an audit trail.

Reasoning-driven optimizer for RAG pipelines: LLM agents diagnose retrieval failures and propose the next config.

This is a sample messenger bot created with Intelligo framework.

Turn an algorithm prompt into a reviewed, narrated, and delivery-verified five-scene Manim tutorial.

This repository contains my Bachelor's degree final year project. It is a Google colab based interactive Virtual Assistant built using open-sourced libraries.

OpenClaw channel plugin for SeaTalk

Botkit(Web-bot) is an open source developer tool for building chat bots, apps and custom integrations for major messaging platforms.

Community Codex Multi-Agent V2 arrangements: lead agents, specialist TOMLs, orchestration skills and optional hooks.

Supervisor/Lead/Peer governance for multi-agent coding — one policy core enforced identically on Pi and Claude Code, brief-scoped authority, exact-SHA independent review.

Interactive subagent orchestration in dedicated tmux panes with fire-and-forget steering, live status widget, and sandboxing for Pi coding agent.

授业 · Book Learning Tutor — Turn any book on your drive into a personal AI tutor (Agent Skill)

A curated list of code world models: executable dynamics, programmatic simulation, and coding agents that build and reason about worlds.

Browser console and diagnostic chamber for Oh-My-Pi (omp) agent sessions — live timeline, steering, real PTY, git/diff/wiki panels, and a panel-plugin platform. Elysia + Preact on Bun.

Create, refine, evolve, and consolidate skills with minimal guidance.

Scaling Training Environments for Long-Horizon Program Reproduction Agents

Error gates for AI coding agents: detects recurring tool-call failures and promotes them into enforced gates - remind first, hard-block on same-session repeat. OpenCode plugin + hook CLI for Claude Code, Codex, Gemini CLI, Cursor, Copilot CLI, Crush.

JevGPT: Thinking Fast and Slow in Autonomous Research

An open-source recreation of AI Co-Scientist: a multi-agent scientific research harness.

A spatial workspace for directing AI coding agents like an engineering director

A secure way to run enterprise AI agents. Building blocks with per-action authorization, prompt-injection containment, least privilege, and a tamper-evident audit log.

SciPIvot: Benchmarking Principal-Investigator Judgment at Pivotal Moments of Scientific AutoResearch Agents

一本开源的 AgentScope Java 中文教程：7 篇 44 章、839 页成书体量的完整 Markdown 源稿，由浅入深覆盖从第一个 Agent 到生产级 Agent 平台的全部主题。

A modern and functional chat assistant based on Local LLM that streamlines internal processes. Weather, internal information, support requests, document uploads, and more, all on one screen!

AI-Assisted Model-Driven Methodology & Low-Code Platform for Serverless Software Development

Real-time two-way relay between WhatsApp groups and Discord channels up to 10 pairs, sender names and attachments, built on discord.js and whatsapp-web.js.

A rule-based travel assistant CLI in Python with sentiment analysis, session memory and live currency conversion.

One prompt → a swarm of 150+ specialist Cursor agents that plan, build, test & harden your whole project in parallel, sharing an Obsidian brain.

TypeScript AI agent runtime. Write your agents in code and choose how predictable they are: deterministic steps where they can be, an LLM only where a step needs judgement, all on your machine.

Observe AI agent runs and control consequential actions. Signed policies, exact-action approvals, AuthZEN API vetoes and evidence. MCP gateway today; connected-channel supervision and scoped reactions toward 1.0. Experimental alpha.

Open-source delivery system for building commercial software with AI agents: governed project board, independent review, security by default, from idea toward release-ready.

@mentions for Codex, Claude Code, and your terminal. A Herdr plugin that lets agents in different panes work together.

Hermes plugin: one /crew message becomes verified work on the kanban board. A coordinator owns every card until its proof passes.

Agentic FUSE filesystem for LLM agents. Structured file operations with JSON feedback, audit logging, undo support, and semantic search.

基于 FastAPI + RAG 的多模型智能客服平台。 装完依赖点一下运行，剩下全在网页上配 —— 不改配置文件，不跑初始化脚本。