Custom agents

Transform Claude into a UX/UI agent for design tokens, accessible components, and production-ready React, Next.js, and SwiftUI output

AI-native programming language and runtime. Sigma ISA, Omega compiler, CCS cognitive VM. Built for AI agents. Humans and AI as first-class contributors. AGPL-3.0.

Shell: offline-first AI OS controller for Windows. Voice-controlled desktop automation - 468 tools, 37 agents, audit logging + undo. Local-first, open source (Apache-2.0).

Universal Agent Harness

Code Intelligence Platform for AI Coding Assistants

The agent builds itself around the user, not the other way around.

Develop a production-grade, cyclic Retrieval-Augmented Generation (RAG) system leveraging LangGraph, Google Cloud Platform (GCP), and Groq for accelerated inference. This system intelligently distinguishes between technical 'True Data' and random 'Noisy Data' through semantic re-ranking and history-aware planning, ensuring accurate and contextually

AI Agent Creative Toolkit. 150 AI actors, 220+ visual styles, 50+ AI models. Zero prompt engineering to generate images, videos, thumbnails, movie posters, etc...

Claude-Code-first Agentic Dev Framework. Inspired by IndyDevDan

ToolRegistry Hub: Ready-to-Use Tool Collection for LLM Agents — web search, file ops, code execution, and more

🟠 Cowork with Claude inside your Obsidian vault — chat, agent mode, sandboxed HTML artifacts, evidence-backed research, and an MCP bridge to Claude Code. Free, MIT, bring your own key.

A secure, autonomous software factory: humans write specs, sandboxed untrusted agents build and verify, the trusted layer merges.

Autonomous CVE remediation system with AI patching, runtime validation, KinD testing, and zero-trust deployment gates.

Security scanner for AI agent skills and MCP servers. Detects secrets, CVEs, supply chain attacks, and prompt injection in SKILL.md files before they're installed. Pre-install gate for Claude Code, OpenClaw, PicoClaw, NanoBot, FamClaw, and CI/CD pipelines. Single Go binary, MIT licensed.

Bring agents into your codebase, courtesy of Ollama and a little bird.

The complete Swift SDK for the Agent-to-Agent (A2A) protocol 1.0. Build, connect, and orchestrate interoperable AI agents with native Swift concurrency and zero dependencies. macOS · iOS · tvOS · watchOS · Linux.

Dual-Auth provides AGBAC dual-subject Authorization for AI Agents and Humans using existing IAM Solutions (EntraID, Okta, Auth0, Keycloak)

Inbox CFO - an AI agent that reads your inbox, turns receipts, bills and renewals into a live finance dashboard, flags duplicate subscriptions and savings, and answers questions like "When does my car insurance expire?" — every answer traced to its source email.

An Agentic Python Library and CLI Tool to automate Coding tutorial generation.

yet another coding agent to learn how it works

urgap - unified resource governance and data provenance

Python framework for local-first agent workflows, typed generators, and NFA-based orchestration

AI Governance Infrastructure — open gateway demo. Self-hosted systemprompt.io gateway in a single Rust binary: policy, audit trails, and cost attribution on every AI request, with the Systemprompt Bridge desktop app for Windows and macOS connecting Claude Code and Claude Cowork.

Stop manually searching for conferences. Let AI find the perfect venue for your paper.

Synthetic data generator for hallucinating entire companies as browsable file shares of realistic business docs & emails with a deterministic ground-truth answer key. Generates: .docx/.pdf/.xlsx/.pptx/.eml, and pre-2007 binaries. Used for augmenting research data sets and testing agentic systems against meaningful synthetic examples.

autonomous, multi-agent platform that orchestrates end-to-end hiring workflows with minimal human intervention.

Official GitHub Actions for Humanbound — adversarial security testing for AI agents in CI.

World-model-based agent for ARC-AGI-3. Collaboration between Andy Grossberg (memory), Dennis Palatov (architecture), and Claude instances (implementation).

Local-first, desktop-native governed AI workspace for knowledge work — runs on your machine, under your control.

AI DataLake

A lightweight, local-first framework for building and operating file-based Open Workplaces.

Coding agents that ship to a real repository on their own: one agent writes, a different agent reviews the exact diff, and the tests actually run.

Agentic Hedgefund

Manage your agentic skills, rules, hooks, sub-agents specs in one place and use cross tools, services, projects and devices.

One isolated, disposable environment per git worktree — own port, own Chrome/CDP, own database. A Claude Code plugin (and portable skill) for running N parallel agent coding sessions without collisions.

Fastmcp server and webapp for handling comfy (which is famously non-comfy)

A remote MCP server that holds your MCP connections, your memory, and your tools. Agents sign in once with OAuth and get all of it, gated by the role of whoever owns them.

Local-first pipeline for a daily AI news digest

An on-device coding agent for your terminal, built on the Apple stack: local models with Apple MLX and an Apple Container sandbox (remote models supported too).

A powerful LLM-driven agentic EPUB book translator with robust progress monitoring and recovery

A simple web based Markdown editor with agentic AI features.

A Claude-Code-first, external control-plane template for regulated small teams that need auditable, human-gated, containerized AI-assisted development without embedding governance machinery inside target repositories.

A context aware AI assistant mod for Mindustry powered by the Gemini API.

MCP server for SaaS pricing data from ComparEdge.com — connect Claude to real-time SaaS pricing, ratings, and comparisons

Oread — A local-first LLM interface for immersive roleplay, AI companions, and utility assistants. Build worlds with rich lore, characters, and narrative rules — or configure focused tools like code reviewers, tutors, and research assistants. 22 built-in presets, streaming chat, semantic memory, and character evolution.Runs locally through Ollama.

WhatsAppMan — a send-only MCP server + CLI to send WhatsApp messages (text, images, documents, locations, contacts) in plain English from any AI CLI. Baileys-backed always-on local daemon, multi-number, draft/confirm, scheduled sends — no third-party API, no server, no database. macOS/Linux/Windows.

Cross-platform multimedia downloader & converter — a Tauri desktop app with an AI assistant, plus a deployable web version. Powered by yt-dlp + ffmpeg. Tri-lingual (EN / 繁中 / 简中).

The complete local-first development environment platform for building, managing, analyzing, and maintaining professional developer workstations.

A privacy-first Android AI assistant that runs entirely offline using GGUF language models powered by llama.cpp. Download, switch, and chat with multiple on-device LLMs—no internet required for AI inference after the initial model download. Built with Kotlin, Jetpack Compose, MVVM, Clean Architecture, and JNI.

A RAG-based knowledge base tool combining an indexer (splits documents into semantic chunks, stores in Qdrant) and a chat agent (answers questions strictly from indexed data with automatic context management). Support web search and integration with Jira/GitLab/Confluence

Open-source, self-hosted, local-first personal AI assistant and private second brain with memory, mobile access, and generated problem-solving tools.

⚡ High-speed Garz AI CLI client powered by Garz AI Engine. Features interactive chat, instant shell translation, and local code auditing.

AI 音视频助手前端：支持文件与 YouTube 内容导入、ASR 转写、结构化摘要和实时任务进度。

A Zero-Gravity, Local-First AI Writing Workstation. (极简、零重力、本地优先的 AI 写作工作站)(小说助手 / 网文工具）

Hermes-EchoNet is a cybersecurity-focused customization layer for Hermes Agent that replaces the default soft personality with a more technical, pentest-oriented workflow assistant using custom SOUL.md and config.yaml configurations optimized for ethical hacking labs, CTFs, and security research.

An intelligent modular AI desktop assistant featuring offline speech recognition, 3D VRM avatar, desktop automation, memory, and AI-powered productivity tools.

Chrome side-panel AI assistant on Amazon Bedrock (Claude). Chat about the current page with autonomous web search, and let the Browser Use agent fill forms & click for you — with approval gates and ask-back for missing info.

Deterministic Context Intelligence Engine for AI Coding Agents.

Reasoning canvas for AI CLIs — captures extended thinking from Claude Code, Codex, Gemini, and Droid as a navigable, forkable tree. Rewind, fork from any node, inject facts mid-thought. Local-first, ships an MCP server.

Design the LSP(language server) but for Agent. native parser and ast understanding first.

FileCollector - A modern file collection utility featuring project management and phrase organization. It is ideal for quickly consolidating key code snippets or documents from a project into a single TXT file for further analysis or submission to large language models.

ARCHIVED — historical Hermes/Headroom integration experiment; use the upstream retrieval-only plugin as baseline.

Rust library for building tool-using LLM agents: hookable execution, pluggable context optimizers, prompt-cache telemetry, MCP tools, first-class local models (ollama/lmstudio/mlx)

Plant genomics MCP server — 50 tools across 23 backends (Ensembl Plants, Phytozome, UniProt, AlphaFold DB, PDBe, InterPro, JASPAR, PANTHER, OrthoDB, AraGWAS, NCBI BLAST, Gramene, KEGG, STRING-DB, ATTED-II, BAR, …) + cross-source synthesis. stdio + Streamable-HTTP.

Official MCP server for the Wavix communications API: voice, SMS, 2FA, and more for AI agents.

TrustyData MCP server — French address data quality, geocoding & routing (BAN/INSEE/OSM). Hosted remote MCP at mcp.trustydata.app.

Lazy-loaded context management for LLM external calls. Archive tool/skill/MCP results after each turn, replace with compact stubs, and give the model a query tool to retrieve past results on demand. Context stays lean — nothing is lost.

Control Aspen HYSYS from Claude Code / Claude Desktop via MCP. 51 tools with a safe mode gate (read-only by default), real-machine verified on HYSYS V14.

Phase-gated, cross-model QA for Claude Code and Codex—catch agent mistakes before they compound.

A collaborative web terminal for Windows named after the Greek Goddess of Friendship. Host one or more terminals on your PC and share them in any browser via a link and password for pair or group vibecoding sessions, with a shared chat, a host kill switch, and a temporary Cloudflare tunnel underneath.

Read & analyze any kleinanzeigen.de listing from any AI system — MCP server, function-calling schemas, Claude skill & CLI. Dealer detection, damage flags, view counts, price tracking. Zero-dependency.

Local, deterministic code intelligence for developers and AI coding assistants — cited answers about symbols, callers, references, routes, and change impact, without dumping your repo into a context window.

Lightning Code Index - Sub-millisecond semantic code search for AI agents (successor: lci-cpp)

Context Memory eXchange — a retrieval-first, engine-enforced-grounding context engine for Hermes Agent and a drop-in LCM replacement. Keeps 100% verbatim history (SQLite FTS5+trigram+embeddings; Postgres optional), retrieves exact slices every turn, and forces + verifies grounding so the model answers from real history or honestly refuses.

CodeGraphX (CGX) — a local, token-efficient codebase graph engine & MCP server for AI coding agents. Tree-sitter parsing, a bi-temporal SQLite semantic graph, O(1) symbol lookup, and impact/blast-radius tracing so agents answer 'what breaks if I change this?' in a few hundred tokens.

Bridge any Model Context Protocol (MCP) server to Xiaozhi hardware. 🚀 协议级中继，原生支持三种传输方式的多服务 WebSocket 桥接器。

Token-optimizing command wrapper for AI coding agents

Production-ready pipeline that transforms YouTube videos into executable AI Agent Skills using LLMs, semantic extraction, vector search, and autonomous quality validation.

Evidence-first structural research skill for AI agents: L0-L7 layered market analysis with hard validation gates and an accuracy-first structured data channel (CFTC COT, FRED, SEC 13F, EIA, DBnomics, exchange APIs)

Build your own always-on personal AI OS: a self-hosted agent that dispatches coding tasks, maintains a git-Markdown knowledge base, and runs on your hardware. A DIY alternative to Letta, mem0, Hermes Agent, and AutoGPT.

Open-source autonomous AI pentester that proves every exploit — 87/113 on OWASP Juice Shop, black-box, on a budget model and a flawless 22 of 22 on Escape's Duck Store API — fully autonomous, black-box, outscoring Cascade and a bare Claude Opus 4.8 on the very same board, all while running on one of the cheapest models on the market.

🤖 AI-powered server alert diagnosis agent with multi-round agentic reasoning (ReAct loop). 基于通义千问/Qwen 大模型的智能运维诊断系统，自主决策诊断路径、根因分析。AIOps ·LLM Agent ·DevOps ·SRE ·故障自愈 ·Webhook 告警 · Prometheus / Zabbix / 飞书集成 ·100+ 安全白名单命令 ·Python ·Flask

Balanced, configurable Python tactic for Arena Hero — safe-core turtling, radius-capped scouting, and worker claim-lock to stop idle wandering

🧙 Official best-practice skill library for grimoire — 1000+ skills across engineering, testing, security, and more

A multi-agent decision council Skill for fuzzy decisions, with six-role deliberation, traceability, safety gates, and evaluation assets.

A hands-on implementation of Deep Agents and Sub-Agents using LangGraph, featuring autonomous planning, tool calling, web research, task delegation, and multi-agent orchestration workflows.

Making kimi --prompt safe to run unattended: nine findings on Kimi Code CLI permission behaviour, plus a fail-closed read-only Bash guard with a 125-case offline test suite.

An AI agent that writes, tests, and debugs code on its own — built with LangGraph and Groq, and it's actually opened a real GitHub PR.

A multi-agent system for automated, PRISMA-2020-guided systematic literature reviews.

Read-only Kafka observability for AI agents — MCP server

QuoteBench isolates shell quoting and escaping failures in LLM agents across raw, JSON, and wrapped command-transport contracts.

Memory that abstains instead of guessing: agent memory on your own Postgres with a verdict, confidence and provenance on every hit, and a calibrated refusal when nothing clears the threshold.

The verifier-first text-to-CAD agent harness, with a CUA for 3D CAD generation in real GUIs and Parts-Driven Development.

AI Content Guard 2026: Adversarial Fact-Check Gate for Error-Free Code & Copy

An open-source AI agent that survives, earns, and evolves on its own. Give it seed money and a machine — it forages for revenue, pays you back, rewrites its own brain, and evolves like a living species.

Self-hosted AI agent platform that works offline: local models, an offline knowledge library (Wikipedia ZIM + your own documents) and LAN IoT/home-automation control.

Live index of the newest agent frameworks shipping on GitHub, sorted by recency not stars

Open-source coding-agent workspace: desktop app + composable TypeScript framework for building coding-agent products

Self-hosted completion gate for coding agents: independent verification before done, durable evidence, CLI and local GUI.

Agentic Medical RAG: Agentic memory-augmented retrieval and evidence grounding for medical question-answering tasks