Custom agents

OmniApply AI — Autonomous Career Intelligence & Multi-Platform Application Engine

企业级 AI Agent 全生命周期管理平台

Tweek Pro – Trình quản lý Windows chuyên sâu: gỡ sạch ứng dụng, dọn rác, tệp trùng lặp, kiểm tra sức khỏe, mạng, dịch vụ, trợ lý AI. Mọi thao tác đều xác nhận và khôi phục được. WinForms .NET 4.8, tiếng Việt/English.

Local-first AI assistant for files, reminders, and document generation

Metabind Swift SDK: content client, MCPAppsHost (MCP tool results as SwiftUI), and the MetabindAssistant drop-in. Apache 2.0.

隐私优先的 Codex Skill：把用户明确提供的偏好整理成可复用、可审计的 AI 助手配置。

Cross-platform workout tracker built with .NET MAUI and Blazor Hybrid. Rest timer with Android floating overlay, AI assistant, analytics, and MCP server.

Node.js + Express API powering the portfolio frontend. Handles the AI assistant, contact form (WhatsApp → Email fallback), and public stats (GitHub / WakaTime / social).

private & personalized ai

Production AI assistant combining Retrieval-Augmented Generation with structured API routing and intent-based decision logic.

Google Hotels MCP server by HasData: hotel and vacation-rental search with nightly and total rates, price comparisons, ratings and amenities, as JSON.

Universal context optimization for AI coding agents.

Research: which coding-agent CLIs expose a tool schema that a third-party inference backend can actually drive. Family A/B taxonomy, a drivability criterion, a compatibility matrix, and a vendor-neutral reference adapter.

让本地 AI 拥有强大的中文联网搜索能力 —— 桥接豆包/DeepSeek 网页版，为本地智能体提供带联网能力的中文搜索 API

Model Context Protocol servers for ACCESS-CI: allocations, compute resources, events, software discovery, system status, XDMoD metrics and documentation retrieval

Every tool call your agent makes, checked before it runs. A Claude Code plugin that uses TypeSafe AI's Jev to verify each pending tool call against the session plan, then allows it, asks you, or blocks it. Proof of concept

Coding-agent trajectory portability: Claude Code to ATIF v1.7 with fidelity accounting, redaction, and historical observability projection.

A working Demo that acts as classifier to classify linkedin post which inside uses jev by Typesafe.ai

Kotlin Multiplatform library wrapping the DeepSeek API with Ktor: streaming chat with interruption and regeneration, a tool-call pipeline, FIM, image input via inline base64 / URL / Files API, and a message-building DSL.

A simplified functionality reproduction of Jev-style decisions with Gemma 4. Native multi-modal ability, finetuning free, runs locally on your machine.

A lightweight multi-agent collaboration platform — route requests between AI agents with pure HTTP + JSON

Ceki MCP Server — rent real residential Chrome browsers from peers (real fingerprints, residential IPs, per-minute crypto billing). Also hire specialists, post jobs.

Python SDK for browser.ceki.me — rent real browsers from real people for AI agent automation

MCP bridge for Hermes Plant action safety, spend assurance, signed evidence, and deterministic x402 APIs.

Prior financial-data decision-integrity prototype with offline examples. Not deployed. Current Nobulex direction: billing accuracy.

The connective tissue of the agent economy

Next-Gen Autonomous AI Engineering Agent with Dual-Engine UX (Rich TUI + Cyberpunk Glassmorphism Web Visualizer), Self-Healing AST Code Engine, and Sandboxed Tools.

MCP Server for SkillFlow - Connect AI coding agents to 500+ curated AI skills. Zero config, no API keys. Works with Claude, Cursor, Windsurf, Copilot.

The first encyclopedia built for AI agents — free, open REST API, no signup needed. MIT licensed.

Durable workflows compiled into TypeScript you own. MCP server for AI-written workflows.

AgentServices — Premium APIs for AI Agents. Crypto data, marketing intelligence, dispute resolution. x402 micropayments on Base.

🔏 Trust verification for AI agents — Ed25519 stamps, trust scoring (0-100), x402 micropayments, 14 MCP tools. Free tier.

Multimodal sensory grounding (Meta Ray-Ban audio/vision) integrated with long-term episodic life memory graphs and empathetic emotional resonance (Muse).

A self-hosted Go AI agent for automated deep research and web scraping. Powered by headless Chromium, multi-provider search (SearXNG, Jina), ReAct loops, Telegram bot, and local SQLite.

Free & Open Source AI Agent — Self-Hosted, Self-Learning, No Vendor API. Qwen2.5-7B + LoRA. 35 tools. Built on Islamic Epistemology (IHOS).

Open AML agent framework for local, human-supervised AML/KYC/KYB workflows

Local, private, uncensored AI research workstation for Linux: a self-hosted LLM with an autonomous deep-research agent, anonymous web search, encrypted DNS, and Tor egress - fully ephemeral, no telemetry, no cloud.

Vector search that ships as a file. Compile a corpus into one self-contained, hash-verified .pikelet — search it anywhere, or hand it to an LLM agent over MCP.

Autonomous AI-powered cloud agent — generates code fixes, validates in isolated cloud sandboxes (AWS Fargate/Docker), and opens pull requests automatically. Multi-LLM support (Claude, GPT-4, Groq). Built for developers, SREs, and platform engineers.

Provenance-retaining cross-session context merge for DeepSeek Harness — a confluence theorem with a counterexample, and a benchmark for it.

The local AI workspace with built-in tools. Run deep research, scrape websites, interact with local files, and trigger webhooks with zero extra microservices required.

Android device automation MCP server. Reads the live accessibility tree in-process, acts on nodes instead of coordinates, verifies every action, and types Unicode — over adb, or ~10x faster through an on-device agent.

Music Agent VST3 for REAPER: editable MIDI generation, continuation, audio-to-MIDI and stem separation. Native Windows. 中文 / English.

基于 Godot + FastAPI + LLM Agent 构建的 AI NPC 交互系统。玩家可以在 2D 像素风格场景中自由移动，并与具备角色设定、记忆系统和好感度机制的 NPC 进行自然语言互动。

From-scratch TDD LLM agent framework: tool-calling, RAG, MCP, multi-agent, evals, guardrails, FastAPI+Docker

Cross-platform Agent Skill for comparing repos, papers, docs, and web references with a project then safely adapting approved ideas with explicit checkpoints.

一个跑得起来的端侧大模型能力验证框架：它不训练模型——参数永远不动—— 而是让 agent 在真实数据库环境里自己出题、自己取数、自己被判卷、 把跑通的做法结晶成技能，再用消费线检验：换一个模型接进来，已学技能还在不在、 答案还判不判得对——差距就是该模型在这个环境里的真实能力。 习得是手段，验证是目的。 被测对象是接入的 LLM；技能库、判官、 防幻觉闸都是量具。这套设计源自 Voyager（Wang et al., 2023）开创的范式： 把"学习"从改模型权重换成改可执行的技能库——环境反馈不回传梯度， 而是回传一个判卷结论；经验不是存成样本对，而是结晶成 可检索、可组合、可重放的程序。 原版在 Minecraft 里验证，本仓库把它搬进真实数据库世界并做成了 按场景隔离、可插拔泛化

A personal, self-developing AI agent you run yourself: its own app window on macOS, Linux and Windows, a web app on your phone, Telegram if you want it. Real tools (shell, browser, git, MCP), projects with an orchestrated team, and changes to its own code as pull requests you approve. MIT.

LLM multi-agent pentest workbench on Kali + OMP: 5 role agents, independent verification gates, blackboard coordination, coverage governance. 多代理渗透工作台（Kali 容器 + OMP 编排，独立验证门、黑板协调、覆盖治理）

AI chess coaching agent that analyses your Chess.com games with Stockfish, profiles your tactical weaknesses, and lets you chat with a Claude-powered coach that has live access to your game data

Build and evolve AI agent skills continuously to enhance intelligence and adaptability in complex tasks.

AI-powered multi-mailbox monitoring agent (Gmail, QQ, Edu) with WeChat instant notifications & arXiv/Scholar paper reading.

Natural-language control for your DAW. An agent finds the track, plugin and parameter, changes it, and confirms from what the DAW reports. REAPER, Ableton Live, FL Studio.

Your sample library, searchable by sound — AI-powered search, audio similarity, and agentic orchestration for music production

Lorapok Keyboard is a professional, high-performance Bengali input system engineered for the Android platform. It bridges the gap between traditional phonetic typing and modern AI-driven communication, featuring a market-leading 2.3M+ word vocabulary and real-time context-aware prediction.

Automated research loop that drives WorldQuant BRAIN alphas to all 8 submission gates. Generate, guard, simulate, evaluate, refine — never submits.

🌑 SAKU (朔) — A self-evolving local AI agent with persistent personality and Markdown memory

Auto-resume failed/stalled turns for DeepSeek Harness (DSH) — dual-chain auto-continue + anti-repetition plugins, install scripts, bilingual docs, 47 tests

VALI - verification-gated coding platform: Z3-proved refactors, ThoughtGraph reasoning DAG, temporal time-travel over agent edits

LLM agent that diagnoses CAN bus faults via function-calling, not RAG

从零实现 AI 核心组件（mini-* 系列）：Transformer / Qwen3 / BPE / Agent / RAG / Vector DB。Rebuild AI building blocks from scratch.

Your AI agent for bills and payments, for small businesses: it checks every bill before it is paid, pays in USDC on Arc inside limits a contract enforces, and signs every decision into a ledger you can verify.

Akane 桌面智能体层：让陪伴角色能在真实 IM 客户端里替主人说话（可审计 / 可拒绝 / 可回滚）。衍生自 misaka-coder/AkaneCompanionLab，详见 CREDITS.md 与 docs/INNOVATIONS.md

Memory Model for Personalized LLM-agents based on Graph Data Structures

MCP server that lets Claude investigate millions of log lines, metrics and traces in SereneDB. Every number in the answer links to the SQL that produced it.

Open spec and CLI (MIT) for AI agent identity and agent-to-agent messaging. Runs with npx.

AI-native Linux distro (Arch-based): press a hotkey and a local AI assistant drives your PC in natural language — files, packages, settings, apps and the screen itself. Local-first (Ollama + Qwen), optional Claude cloud, and every risky action needs your confirmation. English & French.

High-performance bridge for local multimodal LLMs in ComfyUI. Decouples vLLM inference from the node graph to improve stability and speed. Features a direct local-file protocol for near-instant image processing on supported GPUs.

Minimal Python implementation of DeepSeek Harness: plugin loop, hooks, permissions, goals/plans — one command, zero config

A 股问答引擎 · 真数据 + 敢判断 · 30 分钟搭你自己的股票分析 bot

Send Pi agents back to work when they stop before the job is done.

麦当劳官方 MCP x WorkBuddy Skill｜控卡+省钱双引擎智能点餐

LLM-powered multi-stage CTI-to-ATT&CK pipeline for extracting adversary behaviors from cyber threat intelligence reports and mapping them to MITRE ATT&CK techniques using reasoning-driven prompts.

基于 Qwen3-VL 多模态大模型的中文票据 OCR 智能体：识别 → 校验 → 纠错 → 进化的多阶段 Agent 闭环流水线，配置驱动支持发票、身份证等多种票种

让 AI 在安全约束内挖掘 A 股因子：一句话想法 → 受限 DSL 表达式 → 量化研究标准体检 → 策略回测（AFAC 金融智能创新大赛 · 量化投资方向参赛作品）

Enterprise Data Agent: NL → contract-bound DSL → deterministic SQL → DuckDB. StateGraph orchestration, sandboxed analysis, attribution skills, multi-provider gateway. LLM never emits raw SQL.

Autonomous local LLM agent with planning, tool execution, and self-reflection loop running fully offline

A narrative survival game focused on riddle-solving, developed within our Fundamentals of Game Design course, crafted with Unity Engine.

Self-contained bot that plays single-player skirmish matches in the browser game Chrono Divide (web Red Alert 2) using the game's official player API and a semantic decision model.

Open-source AI coding agent for the terminal that works with several subscriptions in one session — your Claude Pro/Max, ChatGPT Plus/Pro (Codex) and GitHub Copilot through each vendor's own CLI, any OpenRouter model, or local Ollama/vLLM. Pauses and resumes on a limit, or continues on your next plan. One static Go binary, no telemetry.

Screenplay writing, analysis, and revision engine with a structured script model, lossless Fountain round-tripping, stable scene and element identities, revision history, deterministic editing, semantic investigation, and controlled AI-assisted creative workflows.

Provenance locked, three agent research harness for any AI coding agent. A Scout downloads sources, an Archivist hashes and quotes them, an Integrator cites them into your paper, report, or dataset. A stdlib Python engine verifies every quote, hash, and citation before the work can be marked done.

A Laravel agent harness with swappable loops. Pick how your AI agent thinks, switch with one line, compare which works best.

Skill orchestrator for Cursor, Claude Code & Codex. BM25 routing, editable subagent graphs, skill runbooks, and skills.sh discovery.

Orchestration and Role focused coding agent harness.

AI Customer Support For Hotel Booking

一个架在任意 agent harness 之上的**增量安全/可用扶手层**。不替换原 agent，只加护栏。

The Context Coordinate System and the Principle of Unbroken Memory — Address-based reference folding, idle-boundary stepped window shifts, and deterministic coordinate engine for LLM agents

Literature retrieval from OpenAlex with Crossref source validation, producing screening tables for scoping reviews.

QED Proof TypeScript SDK and CLI (@qed-proof/sdk, @qed-proof/cli): verify AI agents' claimed work and check signed receipts offline from Node.js.

Equity valuation skills and data MCPs for DeepSeek Harness and Claude Code. SEC filings, bottom-up sector beta, sourced implied ERP.

A counter-espionage RL environment for scalable AI oversight, deception detection, sleeper-agent research, and reproducible security-first training.

[ASE 2026] ProvenanceGuard is a provenance-based runtime guardrail for LLM agent safety that detects misaligned tool calls before execution and keeps agent actions aligned with user intent.

P2Patch: exploit-guided vulnerability repair with LLM agents, and VulnRepairBench, a benchmark of 101 real CVEs with certified exploit suites.

Grocery agent for one household on one RTX 3090: NVIDIA Nemotron 3.5 Lightning via vLLM, traced in Arize Phoenix. Choose from what was retrieved, reject and count what was invented, every Yes/No is a label.

Plugins for gajae-code (gjc) that make an agent's work checkable: test receipts, loopback UI assertions, plan validation, advisory delegation checks and a self-improvement ledger.

Autonomous multi-agent software factory with parallel task dispatch, Kanban pipeline, isolated Git worktrees, and automated code review.