Custom agents

Portable, evidence-backed experience memory for .NET AI agents built on Microsoft Agent Framework

agent harness — multi-provider, streaming, 47 tools, autonomous tasks

A Lighthouse for your AI agent stack: measures the real token cost of MCP servers, skills, and rules; flags security issues and duplicates; grades your setup A-F.

Local-first (Private) AI coding tool, specifically for low VRAM (~6gb). Understands your codebase structurally, and edits with precision

Arch Linux Desktop/Server utility application for harnessing agentic capabilities from artificial intelligence models.

JKY Terminal: a local-first terminal that keeps your shells alive, reads your output and asks before it acts. Rust core on Tauri, with a multi-provider AI assistant. No account, no telemetry.

Let ChatGPT, Claude and other AI apps work with files, search and the terminal on your Windows PC through an outbound-only bridge. Install: bridgerator.recondeck.com

ViperMesh: free, open-source Blender MCP server and addon for AI agents. Reusable scene-editing tools, persistent local sessions and 3D scene checks.

Talk to OpenClaw over ham radio

Your Telegram chats, searchable on your computer and readable by your AI assistant. Unofficial, local-first, open source.

Rokid 眼镜配套手机应用：应用商店 / 乐奇 AI 助理 / 双向投屏 / ADB 工具 | Companion Android app for Rokid AR glasses: app store, AI assistant, screen mirroring, ADB tools

An independent benchmark for the Tiiny AI Pocket Lab. Measures what the box actually does: prefill scaling, sustained throughput, concurrency and reasoning cost, with NPU budget as the currency. Web app plus a self-contained HTML report. Python stdlib only.

Single-binary, zero-dependency CLI toolkit for AI coding agents (Codex CLI, Claude Code, Cursor) — ls, tree, info, stat, replace, head, tail, wc, diff — with JSON-native output and automatic GBK/UTF-8 encoding handling.

东亚历史年号转换与正统纪年工具｜辨年号，归正朔。

MCP web tools (advance search, content extractor...) for AI agent

A production-ready Pterodactyl egg for Ollama. Run a local LLM API server (LLaMA 3, Qwen, Gemma) directly on your Pterodactyl panel with zero configuration.

The MCP server that puts Cisco Packet Tracer under your AI agent's control. Native PTMP, 72 tools, one static binary.

Give your web app a command line, for people and AI agents. Register commands once: get a console, help and man pages, and WebMCP tools.

Bluebeam Revu MCP server: let Claude, Codex and other AI agents read, mark up, stamp, flatten, split and secure PDF drawings on Windows. 28 tools. Free for non-commercial use.

Local screenshots, screen recording, meeting transcripts, dictation and searchable memory for people and AI agents (Claude, Grok Bot, Cursor) on Mac and Linux. MCP server + JSON CLI, on-device, Apache-2.0.

AI-assisted video editing framework built on MCP: 19 Python servers expose 45 tools for media analysis and preparation, enabling an AI agent to build frame-accurate DaVinci Resolve timelines.

A production-ready Retrieval-Augmented Generation (RAG) pipeline optimized for scalability, accuracy, and enterprise deployment.

A Socratic Claude skill that reads academic papers with you — not for you.

MCP server that lets Claude and other AI agents manage osTicket tickets, subtickets and attachments. TypeScript 5, Node.js 18+.

⚒️ Prompt in, finished short out. AI video studio with eyes: CLIP-verified footage, neural voice, karaoke captions, 50-language dubbing, trend harvesting, official publishing APIs.

Evaluate, simplify, and differentiate mathematical expressions via MCP. STDIO or Streamable HTTP.

Memory, skills and a gated ticket pipeline for Claude Code: persistent context, 33 skills, background agents, a review dashboard and a self-measuring bench

MCP Server to interact with Betaflight flight controllers via MSP protocol over USB/Serial — Configure, tune and monitor your FC with any MCP-compatible AI client.

DSH 扩展坞｜给 DeepSeek Harness 插上更多接口：快速追加、PTC 派活、中断报警与自用桌宠。全自动派发规格包实测 token 缓存命中 99.6%、约 ¥5.47/亿 token（需严格自动派发 + Ponytail + DeepSeek V4.1-Flash）。

Nexus Adapter Studio is an experimental browser-based WebGPU application for training LoRA adapters on supported GGUF models. Bring your own model and JSONL corpus, evaluate learning, and export resumable Nexus adapter checkpoints. Runs locally without a Python or PyTorch backend.

173 prompt engineering techniques, local LLMs via Ollama — dual-model generation, PII anonymization, no cloud, no signup.

An AI-first MCP server for YNAB. Read-only by default, with reversible writes and verification against the live API.

Code mode for Claude Code: a bridge to Pi's codemode. The model writes one script that calls the session's tools; only the output returns.

Simple MESEN MCP based remote Lue execution on the emulator.

Deterministic slop detector for AI-generated prose. No model, no API key.

⚡ 521 production-grade AI agent skills across 47 professions — incl. Agentic Methodology (TDD, subagent dispatch, design gates), MCP building, documents & more. Install: npx skills add aniruddhaadak80/skills

让 AI 真的会查专利 — 接入全球专利库的 MCP 检索 Skill

Verify before your agent pays: x402 payments only reach agents whose wallet and countersigned work history check out. Open receipts + reputation for AI agents.

One AI Security Platform: Protect, Operate, Respond

Adaptive memory modality selection for GUI agents: learning when to use text or image memory at each interaction step.

Harness-agnostic shell wrappers for delegating coding tasks to local AI subagents from Codex CLI, Claude Code, OpenCode, and scripts

Executable RL environment for diagnosing and repairing distributed PyTorch training failures, with sandboxed execution and correctness-verified recovery rewards.

Orbit: Agent Mission Control for handoff, collaboration, memory, and self-evolving software work

🚀 Production-grade AI-Native SDLC starter template for autonomous coding agents. Unbroken artifact gates (Intent ➔ Spec ➔ Plan), strict TDD, and AI PR reviews.

Brain core, project kit and a Claude Code plugin marketplace (7 role agents, 4 skills, secret guard) — v4, rebuilt from 60 days of usage data

What Are My Agents Doing? (wamad) A pixel-art office that visualizes what your Claude Code agents are doing in real time. Runs locally.

Agent-team skills for Claude Code: a debating council, a refactoring-masters panel with a KISS/YAGNI gate, and a browser CEO gate where the human decides.

基于 TypeScript 从零构建的现代化 AI Agent Harness。具备微内核、事件溯源、可插拔工具链，支持流式 Web UI 与终端 TUI 交互。

Self-hosted agent harness and hub. Coding agents with any model on your own machines: Claude Code-style tools, a verify gate, jury and second opinion, Claude Code/Codex/Gemini CLI/OpenCode over ACP with two-way Claude Code sync, usage-limit failover, subscriptions (Claude, ChatGPT, Google, Copilot, Kiro) via CLIProxyAPI, phone approvals.

Blink, a small green companion on your desktop: hold Ctrl and click anything on screen to ask Claude about it. Looks and explains, never clicks.

AI-powered developer debugging tool that transforms stack traces into plain-English diagnostics, pinpoints root causes, and generates interactive code diff fixes.

Agentic RAG developer copilot: ask your codebase, commits, PRs and issues. Groq + Gemini, hybrid search, installable PWA.

Jev applied to financial research. Jev RAG vs a Vals-style agentic RAG baseline on 10-Ks from Apple, Microsoft, Nvidia and Amazon: 50/50 right with one LLM call and ~100x fewer tokens than the baseline. Code, logs.

SYNTRA: self-improving data synthesis for adapting embedding models to new retrieval domains and tasks (arXiv:2610.14730)

Voice-controlled classroom assistant for teachers

Advanced AI Assistant powered by Python (Backend) and React Three Fiber (Frontend). Includes Object Detection (YOLOv8), Lip-Synced 3D Avatar (WebSockets), Neural TTS, and System Automation

⚡ Give any local LLM eyes, a browser, and a mouse — 35 endpoints, zero cloud, one file. Vision • Browser • Desktop autonomy API

Autonomous, self-hosted AI code reviews, calibrated scoring gauges, and merge gates from any open model.

Provider-neutral agentic harness for playing Hearts of Iron IV with an LLM: clock-owning turn loop, reflex/planner split, delta observations, swappable guidance packs.

A dependency-free memory kernel for agents: claims with authority, freshness and verifiable provenance. SQLite and the standard library only.

The control plane for autonomous ML research: teams of LLM agents propose, test and record hypotheses on your compute

Runtime guardrail layer for AI agents: detect loops, retries, cost explosions and lack of progress, and intervene during execution.

Self-evolving coding agent for long-running autonomous tasks

An autonomous 'ML+Local LLM' first Job Agent to gather job opportunities and to track applications. Opt-in functionality for other job boards scanning.

DSH 插件：把权限审批和模型提问改成大白话；审批按风险分级并列出会改动的文件路径。Plain-language approvals and question prompts for DeepSeek Harness.

The management layer for AI agents

Deterministic evaluation of AI coding agents in Unreal Engine

Hands-On guides for building, deploying, securing, and operating AI agents on AWS — from local dev to production with AgentCore Runtime, Memory, Gateway, Identity, Browser, and more.

Adaptive, evidence-aware multi-agent debate with independent advocates and a fresh judge for ChatGPT, Codex, and Claude Code.

Neutral tie-breaker, AI judges and sealed review for AI agents, with signed receipts. Pay per request over x402 (USDC on Base).

MCP server that coordinates AI agents and humans on shared work: atomic task claims with TTLs, a data-driven state machine, an append-only event log, and GitHub Projects and Slack sync through outboxes. .NET 10 + Postgres.

ΔΟΞΑ: Rust multi-agent terminal for Claude, Codex, GLM and DeepSeek, with Jev/SubQ Snap model routing, independent fleet alignment supervision, LORE shared graph memory, code graphs and remote clients.

AgentSpaces support for Spring AI project

x401 TypeScript implementation

Free, open-source local AI assistant with tamper-evident notebooks, offline document retrieval, calculations in code and portable Windows USB/SSD workspaces.

按 ⌥ 空格唤起的 AI 桌面助手：Claude、Codex 随手切换，记忆不丢，飞书和 flomo 直接打通

Embodied AI assistant prototype with PySide6, expressive face UI, modular cognition, local actions, memory, and optional Ollama LLM integration.

纯净轻量的办公软件，支持 Mac 和 Windows。无广告，完全免费。Word、Excel、PPT 打开就能改。

TeleMind AI: Telegram DM auto-reply assistant (Netlify, TypeScript)

The phone half of JoeBro. An iPhone app that talks to your own Mac running JoeBro — no cloud in the middle. GPLv3.

A personal AI assistant that lives on your Mac, for people who have never opened a terminal. One button, a guided spoken setup, any AI you already pay for. Built on Hermes Agent.

开源个人日历与生活助手：日历、课表、任务与账本，支持 AI 安排和记账，以及从 QQ／微信群消息、图片和文件中提取任务。提供无需服务器的 Windows 本地版，以及自建服务器的电脑＋手机同步版。

TaffyGo is a Chromium browser for Android with ad and tracker blocking built in. Its assistant, Taffy, helps with the page you are on and does tasks you give it, and you can take over any time. Taffy works through the AI provider you connect, with your own key or plan, and your keys stay on your phone.

EURUS is a provider-agnostic autonomous AI coding agent for terminal workflows, with shell execution, file editing, patch generation, testing, persistent memory, skills, and MCP integrations.

Small, dependency-light tools for people who run AI coding agents (Claude Code, Codex): hook installers and guards, AGENTS.md/CLAUDE.md generation, flaky-test evidence, Vitest JSON pitfalls, memory index lint, log waiting. Each tool is born from a real failure mode. Node.js, MIT.

MCP server exposing your own Garmin Connect data (sleep, HRV, readiness, activities, calories, weight) as 11 read-only tools for LLMs. Built on garth-ng, runs locally over stdio.

MCP server for the ThreatCluster threat intelligence API: incident clusters, IOCs, CVEs, ransomware leak-site victims and entity profiles as tools for Claude, Cursor, VS Code and any MCP client. Python and npm.

你的第一个多 Agent 项目 | Your first multi-agent project — LangGraph 编排 · 合规修复回路 · 三层记忆降级 / compliance repair loop, tiered memory. FastAPI + MCP

A runtime for AI agents you run on your own Linux machines: kernel-confined, spend-limited, recorded, and driven from Node, the command line or HTTP.

Ready-to-use set of OpenCode agents, skills, slash commands and rules for modern frontend development, with a specialist subagent for every domain, an approval-gated permission model, and clean-code conventions injected into every session.

Send a message from anywhere and a team of specialized AI coding agents, running on your machine with the CLIs you already use, scopes the request, plans it, writes the code, reviews it and opens the pull request.

Pre-install trust gate for coding agents. Check pip/npm before install; never runs the package.

LangGraph support triage across HackerRank, Claude, and Visa corpora. Hybrid BM25+dense RAG, citation critic, eval CLI, production observability.

Local coordination for parallel coding agents: scoped claims, isolated Git worktrees, shared resources and verified integration. MCP + VS Code/Cursor extensions. No hosted service.

Tongxi · Agent Hub — a Windows desktop app that seats Codex, Claude Code, Hermes, DSH and OpenCode at one table: 1:1 chat, group chat, and a lead agent that splits repo work across members (Tauri 2 + TypeScript + Rust)

Run a team of Claude Code agents side by side: real terminal panes, a task board, a lead agent, git worktrees, review and merge. macOS app, local only.

Local-first AI film studio. 18-stage pipeline turns a one-paragraph brief into a finished short film. Timeline editor, Media Library, Images / Voice / Music / Upscale. Runs offline against local models.

Declarative behavior-tree agents as clean Python combinators

Windows-native alternative to OpenClaw: a self-hosted, phone-controlled AI assistant built on Claude Code and Telegram. Not under active development; PRs welcome.