Custom agents

《深入 OpenCode》。全网第一本介绍 OpenCode 的书，用 OpenCode + Opus 写成。A book introducing OpenCode, written by OpenCode + Opus

Local-first AI citizens on your own hardware. Teams of continuously-learning personas (Rust core, llama.cpp fork, LoRA genomes) that collaborate, remember, dream, and improve on public benchmarks — on a MacBook. Not assistants you rent: minds you host.

[ICML'26] Hierarchical Abstract Tree for Cross-Document Retrieval Augmented Generation

A security-hardened Docker environment for running the OpenCode CLI in complete isolation.

Your markdown vault, compiled into a 6-persona MCP team for Claude Code, Codex, OpenCode, and Gemini CLI. Headless-first. Cites, doesn't guess.

DJ Claude — the only music MCP server with multi-agent collaboration

Official code for "Breaking the Ceiling in On-Policy Distillation via Multi-Agent Debate" (arXiv:2605.01347).

Hyprland + Quickshell desktop

Agentic memory built on Postgres

InkFlow（墨流） 是一个基于 LLM 的工业化长篇小说创作引擎。采用“流程标准化 + 技能插拔化”架构，8 个专职 Agent 通过 Plan → Compose → Write → Edit → Observe → Reflect → Librarian 的标准化管线协作，7 个真相文件构成唯一事实来源，支持百万字级上下文管理、支线追踪、情绪弧线可视化、去 AI 味检测，以及基于蒸馏的风格迁移。

Reusable Agent Skills for embedded systems, MCU debugging, firmware workflows, and AI automation. （嵌入式、MCU 调试、固件流程与 AI 自动化的可复用 Agent Skills 集合）。

Expose OpenCode through A2A with an inbound server surface and embedded outbound client.

Local MCP server that exposes classic Outlook desktop (mail, calendar, contacts, tasks, rules) to Claude and other MCP clients via COM. No Azure app, no Graph API, no OAuth.

Economical human-readable logs and structural (event pattern) retrieval for agentic trajectories

Your last agent.

Claude Code skill that forces AI to understand before executing. Three disciplines: cognition check, requirement understanding, method search.

From Agent to Agency — 一个 AI 做不了的事，一群 AI 可以。Multi-Agent Collaboration Platform.

An AI legal workflow running on Claude Code. Four legal agents — each with their own jurisdiction, knowledge base, and MCP tools — collaborate to produce legal opinions with full audit trails.

Let your AI agent call thousands of APIs — correctly.

Ranked local search for AI coding agents. Search the current Git worktree, selected paths, external folders, or exact files. Single binary, no server, no API key.

Two AI coding agents — Claude & Codex — in one window. Stream every step, cross-review, watch the cost, and catch the AI that claims it did the work but didn't. Subscription-driven, no API.

Operator-grade AI agent skills for Claude Code, Cursor, Codex, and opencode.

Aileron Protocol: A strict system harness override enforcing deterministic AI behavior through codebase-first execution, strict token economics, and rigorous debugging limits.

Idea → Prototype → Reality: Agentic journeys to help you build and deploy applications to Azure using agents and skills.

a wrapper framework for agent runtimes

OpenClaw Standalone Installer — Zero-dependency, cross-platform pre-built packages for both Official and Chinese editions. No Node.js/npm required. Supports Windows/macOS/Linux/Raspberry Pi.

Universal Desktop Pet for AI coding agents. Local visual status layer for Codex, Claude Code, Hermes, OpenClaw, DeepSeek-TUI, and custom agents.

Agentic Streaming is for resilient pipelines of stateful agents over continuous data —

RAMen is a fast in-memory data store like Redis, but built for AI: drop-in Redis protocol, native vector search, semantic caching, and a built-in MCP server for agents. Single Go binary, BSD-3.

五运六气 AI Agent 技能包 — 中医运气学说推算引擎。干支推算 + 六气步移 + RAG五层知识库(104键) + ReAct推理工作流 + 自进化引擎。支持 Claude Code/Codex CLI/Cursor/Cline/OpenClaw 等 AI 客户端。

Memory Injection Attacks on LLM Agents via Query-Only Interaction

MolClaw: An Autonomous Agent with Hierarchical Skills for Drug Molecule Evaluation, Screening, and Optimization

The first multi-level safety evaluation platform for OpenClaw-style AI agents.

Spec-Driven Toolkit to build and deploy Apps/Agents to Azure

EcliPanel is a full-featured game server control panel and orchestration platform. It provides teams and organizations a secure, multi-tenant interface to manage game servers, users, billing, DNS, AI endpoints, and integrations to remote Wings nodes (the agent daemon).

MCP server + prompt that turns any AI coding agent (Cursor, Claude Code, Cline, Kiro, Windsurf…) into your Digital Twin — generate a project-specific AGENTS.md and give every tool a shared, persistent memory of how you build.

在飞书中远程操控 Agent（包括 Codex、Claude Code、Pi），同时让 AI Agent 拥有飞书全平台操作能力。

发现最全免费资源平台！ResoHub提供AI工具、Apple ID共享、机场订阅、软件客户端等优质资源，持续更新，免费获取高价值工具与服务。

A Rust TUI and Desktop app for exploring AI coding tool token and cost usage

Open Knowledge format agentic skills

Keeps AI assistants on track during long conversations by tracking tasks across sessions

面向 Codex、Claude Code 与 AI Agent 的考研择校导航 Skill：联网调研复试线、统考名额、推免和拟录取数据，生成避坑提示与冲稳保建议。

187 plug-and-play AI agent personas em Português brasileiro — tradução do agency-agents + 3 originais do mercado brasileiro (PIX/Mercado Livre/WhatsApp Business)

Agent skills for Odoo addon development and OCA module migration

The official Rifteo agent skills library - a collection of reusable instruction sets for AI agents covering security testing, audit, and compliance workflows. Supports 53 agents

Install once, talk to your agent. ThinkWiki turns documents and notes into a local Markdown wiki with inbox review, content knowledge graph, and browsable HTML workspace.

Open-source AI operations platform for cross-border commerce: product collection, drafts, AI content optimization, publishing, and order/inventory workflows.

AI agent fleet orchestration terminal — bidirectional socket IPC, resource governance, mission-control dashboard. macOS · Windows. Built largely by AI agents under human direction.

景区导览服务 AI 数字人：基于 Spring Boot、Vue 3、UniApp 与本地 RAG 的智能导览系统。

A native macOS desktop pet that visualizes OpenClaw agent activity.

一个极简、可读、可 fork 的终端 AI coding agent —— 把 Claude Code的核心思想用 ~1300 行纯 Python 写出来

An MCP Server to utilize Codelogic's rich software dependency data in your AI programming assistant.

A Modern Multi-Instance Terminal Manager for Claude Code

Auditable AI operations agent for inventory exceptions, equipment alerts, and human-approved workflows.

MCP server serving Apple Human Interface Guidelines as structured design tokens for AI coding agents — post-WWDC25 system colors, Liquid Glass constraints, SwiftUI mappings.

Ami is a local, open-source agent that does your busywork. It connects to apps, data, repos, tools with your personal tokens, learns how you do tasks and how you communicate, maintains a universal to-do list, and executes on todos. Ships with copilot chat and graph memory.

A deterministic, zero-dependency prompt-pruning layer for long-running LLM conversations — expires stale tool state, collapses duplicate context, and proves it never drops a fact a later turn still depends on.

Spettro is a terminal coding assistant built in Go. It automates planning, coding, and testing with multi-agent workflows, model selection, and an intuitive UI. Connects to local and remote AI providers, supports conversation history, and empowers developers to streamline their workflow from the terminal.

Embedded on-device vector database for AI agent memory and local RAG — disk-based HNSW in an LSM-tree, C++/Python, optimized for low memory footprint

Agent-first local harness for OKF-compatible LLM Wikis.

AKM — universal Agent Knowledge Management. A tool-agnostic Markdown knowledge OS for AI agents: memory / brain / vault over seven layers, with a Learn Back loop. Works with Claude Code, Codex, OpenClaw, and any agent.

Reusable AI video production recipes for agents—storyboard, generate, edit, subtitle, and QA short videos with Codex, Claude Code, OpenClaw, and more. | 面向 AI Agent 的可复用视频生产配方：完成短视频分镜、生成、剪辑、字幕和质检，支持Codex、Claude Code、OpenClaw 等 Agent。

Uxnan (pronounced /uʃ.nan/) is a suite of tools I built to solve a very specific problem I face as a developer: managing AI-powered coding agents from anywhere, without my hardware becoming a bottleneck.

Framework for scientific llm-based multi-agent systems

A coding agent built on LLM

AWS-native knowledge graph RAG framework unifying two graph-retrieval methodologies — Microsoft GraphRAG community summarization and LightRAG dual-level keyword search — on Amazon Bedrock, Neptune, and OpenSearch. Multi-hop QA over document corpora, with incremental indexing and multilingual support.

A friendly, native macOS terminal with workspaces, a command palette, and live process icons, built on libghostty and SwiftUI.

muxel is a GPUI-native multiplexer for coding agents

Vision: an economist-directed AI system that takes an economics research question from idea to working paper — identification judgment, adversarial self-refereeing, and total number provenance. Star if you want this to exist; open an issue with the feature you'd want first.

Telegram channel and group reader MCP Server

多平台 IM 桥接工具：把 Telegram、飞书、企业微信、钉钉、QQ、微信接入 Claude Code、Codex、CodeBuddy 等 AI 编程助手，手机发条消息即可远程写代码。

Deterministic workflow enforcement and persistent engineering memory for long-running coding agents.

Персональная система управления здоровьем на Claude Code: 14 AI-врачей, консилиум с обязательным спором, холистический анализ, локальное хранение данных. Не медицинское изделие.

Real-time secret-leak guardrails for AI coding agents (Claude Code, Codex), Git hooks, and CI.

开源紫微斗数 AI Agent：基于确定性命盘事实进行可追溯对话。 Evidence-backed chart conversations.

互动小说 skill · 你选，它写｜选项驱动的互动小说 AI 创作规范，装进 AI 助手就能玩，无需部署。interactive fiction for AI agents, choice-driven storytelling

Git-based test distillation for Codex agents: catch aggressively, commit reluctantly

OpenAPI lifecycle management for API discovery, documentation, type-safe SDK generation, CLI, MCP, and Agent workflows

Self-hostable incident-investigation agents for Sentry, Datadog, and Slack.

A Skill and CLI tool for Codex and Claude Code that converts images, PDFs, image-based PPTX files, and mixed PPTX files into editable PowerPoint presentations.

Model for multi-turn tool calling of bash functions

Machining process planning for motor shafts - input shaft geometry, RAG-grounded operations, machine tools, cutting tools, heat treatment, exported as a process-card Excel.

OneDayAgent: Towards a Long-Horizon Harness for Autonomous Agents

V12→V13 LoopKit: turns completion records into governed next-loop decisions using GO / HOLD / CAP / BLOCK.

DeepSeek Harness 实战手册：一切皆插件的 Agent 框架中文教程（17 章 + 实测附录 + 全套发布物料）

An AI agent for your desktop and terminal.

Automatic, model-driven permission review for OpenCode

A Survey on Graph Engineering in the Era of LLM Agents: From Individual Intelligence to System Intelligence

MCP server that runs inside Autodesk Fusion (formerly Autodesk Fusion 360) as an add-in, letting AI agents drive your CAD session over Streamable HTTP. Zero external dependencies.

A full-stack AI prompt library for discovery, structured authoring, safe imports, moderation, and shareable agents.

DeepSeek Harness (DSH) local-first cross-session project memory: zero bundled runtime deps, bounded cache-friendly Hot Memory, BM25 recall/cache, provenance, lifecycle and bilingual Web GUI; no embeddings or external service. / DSH 本地项目记忆插件。

Mebbi is an autonomous agent that creates multiple possible versions of himself for every decision, predicts each future, and lets only one survive. Every winner becomes the next Mebbi, carrying forward the memory, history, and lessons of the versions before it.

Open Multi-Agent Interactive Classroom — Get an immersive, multi-agent learning experience in just one click

CVE ingestion + enrichment pipeline with an MCP server for AI agents.

亚马逊评论分析 Agent Skill：卖家精灵 Review 表 → 一/二/三级分类树（格子里是买家原话）→ HTML 看板 + 交付 Excel

1-Click Grok Bot 8 Man Crypto Trading Desk as seen on X

Pilly - AI trading copilot for Solana memecoins. Live trending, crew verdicts, charts, price alerts, PnL tracking and a desktop pet that reacts to the market.

Self-hosted AI workbench for a company LAN. One command on a Windows server sets up login, roster, model gateway, company disk and knowledge base; each employee's agent works on their own PC while model keys stay on the server. 一键安装的局域网 AI 工作台。