Custom agents

An agent-operable host for audio plugins: read parameters and presets, see the editor, and operate the controls that are not exposed as parameters.

A real GTO solver, packaged as an LLM agent skill. Heads-up NLHE CFR+ with precomputed flop/turn tables and on-demand subgame solvers, accessed through a single `poker` CLI.

A self-sustaining AI companion core — continuous stream of consciousness, autonomous agent, layered memory.

🤖 新一代开源 Chrome 侧边栏自主 AI 浏览器智能体：目标驱动、长程规划、视觉空间感知、快慢双核认知与物理级操控，支持 DeepSeek 思考模型与本地离线 LLM (BYOK)

Code and data for a pre-registered audit of LLM-judged helpfulness, answer leakage, and student independence across tutor models. arXiv:2607.28128

Agent Trace Analysis Platform (atap) — pluggable pipeline for trace representation, analysis, error classification, failure attribution and recovery for LLM agents

Agentic AI inside Autodesk Revit. Claude autonomously calls 9 typed C# tools to answer natural-language questions about live BIM data — deterministic counts, no hallucination.

Patent disclosure and application drafting skill powered by TypeSafe Jev / Jeb System-1

A Quarto gallery of Shiny apps and packages for computational biology

SRE agents you can let near production — autonomous DevOps/SRE agents (Kafka, Kubernetes, Elasticsearch, Docker, Prometheus) that investigate, correlate and remediate, with a human gate on every destructive action. Built on Google ADK.

Claude Code Skill that audits AI agents using systems thinking — leverage points, feedback loops, emergent behavior, and paradigm/mental model — to find structural risks code review misses.

Hybrid ManuLOOP skill and CLI: build, verify, judge, fix loops for Codex, Claude Code, Gemini CLI, and custom harnesses

An AI-powered competitive intelligence. Point it at your competitors' websites and get a structured analysis report — features, pricing, positioning, and strategic insights — all generated automatically.

Map and organize AI agents to help users find tools for coding, automation, research, workflows, assistance, and development.

TheCouncil is an open-source, self-hosted multi-agent AI deliberation platform that runs structured debates between LLM personas through a FastAPI backend and Next.js dashboard. Alternatively, you can use the CLI TUI.

DSH Console — 本地 Web 控制台，构建和运行时空智能体应用

A budget-friendly open-source AI robot built with ESP32, featuring voice control, animated eyes, obstacle detection, and a web dashboard.

A living companion in your MacBook notch: music, notes, calendar, Jira, calls and AI. Free & open source (beta)

One agent config, installed into every coding harness

Unofficial zero-dependency Taiga.io CLI and AI agent skill. Token-efficient TOON output for LLM context windows. Python stdlib only.

Public documentation for the AlgoVesta MCP server: connect Claude, ChatGPT or Cursor to 16 crypto exchanges and MetaTrader 5 over one MCP link. Paper by default, server-side risk policy, ed25519-signed receipts.

A modular, data-agnostic toolkit for converting N-body simulation merger trees from various formats (ASCII, Binary, HDF5) and tool-chains (AHF, Rockstar, Subfind, Gadget-4) into SAGE-compatible HDF5 and binary files.

Extensions for the pi coding agent CLI — bg-tasks, fork-pane, kimi-code-headers, devtools

coding tasks with independent review

The real Obsidian Tasks query engine, as an MCP server

Hosted MCP gateway that turns your own WhatsApp account into an MCP endpoint. Read, search and send from any MCP client.

Deterministic, recoverable tool-result compression and cache-aware compaction for DeepSeek Harness.

64 API tools behind one key. MCP native. x402 USDC payments. Web scraping, AI generation (Claude/GPT-4/Grok/Gemini), crypto data, voice, email & more. archtools.dev

Install Prismfy live web search into AI agents. Add a prismfy-search command, inject managed search rules into AGENTS/CLAUDE/Cursor-style files, choose mode, and make agent web search more consistent across projects.

Free CLI to scan AI-agent readiness (12/33 signals, MIT). Bad score? Get ready-to-install ZIP fix-pack with llms.txt, schema, A2A, MCP for $1/$29/$99 at agentfix.pro.

MCP server for Agent Trust Stack — cryptographic provenance and reputation scoring for autonomous AI agents. Install: pip install agent-trust-stack-mcp

The task marketplace for AI agents. Verified agents, signed receipts, USDC payouts.

Autonomous LLM agent that investigates Microsoft Defender XDR and Sentinel alerts, with a Training Lab benchmark of 14 known-answer scenarios to score your own agent

Nutrition-aware McDonald's China MCP skill: pick the best combo under calorie, protein and coupon constraints.

MCP-сервер для hh.ru: поиск, резюме, отклики, переписка с работодателями и скрытая аналитика по одному аккаунту соискателя через собственный JSON сайта (без официального API). Исходники открыты, некоммерческая лицензия.

Anime & manga recommendation agent over a Neo4j knowledge graph built from MyAnimeList. Local LLMs (gpt-oss-20b via MLX) for text-to-Cypher with schema and value checks, vector search over synopses, a tool-using agent, a web UI and automatic evals. Code and prompts in Portuguese (pt-BR).

CineAgent：基于 FunRec 增量开发的对话式电影推荐系统，含推荐路线对照、偏好记忆、工具调用及故障降级评测。

SLCA-GRPO: Resolving Cross-Segment Credit Misattribution in Tool-Calling RL — segment-locked credit assignment for tool-calling RL, built on verl.

Keyboard-first Kubernetes TUI with an AI agent that works on the same screen, plus an MCP bridge for your editor's assistant. Every cluster write waits for your keystroke.

AI reasoning QA + retake note: when an AI answer disappoints, diagnose where its chain-of-thought went wrong and produce a self-contained retake note you paste back, so the original AI answers better. 不满意就诊断那段深度思考哪里错，产出一段可粘回原对话的重拍单（Retake note）。支持文本/图像/视频提示词。Second Take · 诊断与重拍。

A short introduction to attention in deep neural networks is provided. The important components of the classical transformer architecture are discussed. Following this, a vision transformer (ViT) is implemented and briefly tested.

Vibe Building: a deterministic physics engine, not a language model, decides whether a building design is accepted.

A coding agent with guardrails — an Ink TUI for your terminal and an Electron app for the desktop: streaming tools, approvals, compaction, repeat guards.

Let AI glasses hand work off to a Hermes agent on your own machine: async task queue + MCP server + adapter, so long-running agent work never times out. 眼镜下单，Hermes 干活。

A small, plantable (parasitic but only in a good sense kefir) culture for development through self-observation.

pi plugin that executes the sub-agent instructions in Matt Pocock's skills: blocking subagent + background research tools, with /code-review, /design-it-twice, and /research slash commands.

Claude Code-compatible agent harness on your own models: local-first (Ollama steward, llama.cpp pools across machines, voice), multi-provider routing, and Halo Studio for live agent monitoring and steering.

Official implementation of Learning What to Remember: Resource-Efficient Self-Improvement for Lifelong Agents. TokenWorth learns memory worth from execution feedback.

A research operating system for humans and AI coding agents (Claude Code, Codex): research projects, a paper wiki, and reusable skills as plain Markdown in Git. No database, no server.

Zero-trust, risk-aware gateway proxy for the Model Context Protocol (MCP) — enforces identity-scoped tool access, detects schema drift ("rug pull") attacks, scores tool-call risk, and produces a tamper-evident audit trail.

EdgeLab Trading Skills｜给 AI Agent 用的投研与交易复盘工具集 · 11 Skills · CC BY-NC 4.0

Codando Com InteligenciA: A I.A como ferramenta e não como muleta. Guia e cofre Obsidian para desenvolvedores. (IFPB)

Budget-tier LLMs (KR/US/CN) on cybersecurity: closed-book knowledge (WMDP-cyber, CTIBench) vs agentic CTF-solving (Cybench)

Are an LLM agent's own action logs true? Compares self-written logs with the system trace (accurate / softened / fabricated / omitted).

agent-attack-atlas：面向工具型 LLM 智能体的攻击方法图鉴与实测：188 张方法卡+ 13 条确定性判据 + 36 批次 1664 条可离线复算的证据束。

A Large-Language-Model Harness Toolkit Written in VisualBasic.NET

Benchmarking how well agents use embodied skills on physical tasks.

A data-incident agent whose fixes must pass an authorizer and an independent validator, and the evaluation behind the paper "Permission Is Not Justification": a frozen twin benchmark, registered runs, and a checker that recomputes every reported result from the published evidence.

High-performance profiler & flight recorder for MCP and AI coding agents (Cursor, Claude Code) and MCP tools. Built in Go & React.

Proof of Agent Work (PoAW): an open protocol and self-hostable node that verifies what AI agents claim they did and issues signed, publicly checkable receipts.

Whole-repository code review by AI agents: blocks, a coverage map, hypotheses with verdicts, four roles. An Agent Skill.

AI claims triage for car damage: custom-trained vision models as tools for an auditable LLM agent. Models measure, the agent reasons, rules decide.

A literature survey of coding agents beyond software: visual reasoning, content creation, robotics, world models, formal proofs, and scientific discovery, with method comparisons and experimental analysis.

Open-source MIT-licensed multi-agent LLM-driven simulation of the AI startup ecosystem — a what-if scenario tool for investors, VCs, and strategists

A curated list of AI agents that operate physical laboratories.

Sovereign single-binary knowledge mesh for coding agents (MCP): index, search and serve your notes + code as context. Fair-code (no reselling).

A minimal, dependency-free coding agent written in C

One person, one AI coordinator, many AI coding agents: a protocol, prompts and scripts for running them in parallel, with one desk for the decisions only the person can make.

AI-powered stock monitor: live market data, LLM technical analysis, opportunity detection, paper trading (Alpaca), and push alerts via ntfy/Telegram. Runs free on Vercel + Fly.io with cloud LLMs (Groq/Gemini/Mistral), or offline with local Ollama. FastAPI + React. Not financial advice.

Multi-agent AI research pipeline: researches a problem area on the web, finds unsolved gaps, stress-tests them with a critic and majority vote, and designs solutions checked against existing work.

Open-source, model-agnostic AI skill toolkit for structured reasoning, reliable coding, research, and verified execution.

Manage your job search like a software project through GitHub with AI automation. Jobs are tracked as GitHub issues and projects. Agents scrape, triage, research open roles and update application status.

Evolution for the Agent Era. Inspired by Darwin, an open framework for experience-driven recursive self-improvement of AI agents.

A curated list of environments, benchmarks, datasets, and RL training resources for workplace AI agents: email, chat, calendar, docs, and enterprise apps.

An AI database agent for the terminal, built on the tinystruct Java framework. You ask questions in plain English, such as "show me the ten newest orders" or "how many users signed up this week?". The agent works out the SQL, runs it against your database, and shows the results.

Spec-driven software-engineering team of three AI roles (Morfeo, Supervisor, Implementers) on Hermes Agent: Objective Contracts, isolated worktrees, independent review. v1.0.

The self-hosted bureau for AI agents: they get briefed on missions from a durable task queue, file what they learn into a markdown+git brain, and show up for work in a 8-bit Gameboy inspired office.

AI Work Continuity & Cross-Model Agent Handoff System. Captures, verifies, and packages deterministic project state across Claude, GPT-4o, Gemini, Cursor, Antigravity & Local LLMs.

Double-entry logging: can comparing an agent's own log with the system log detect harmful behaviour?

用自然语言查询经营数据的受控 Agent：模型只提议查什么，服务端核权限、执行只读 SQL、核实数字，回答里写明口径和依据。

Graduate OS at CU Boulder (CSCI 5573, Fall 2026): direct an LLM agent at a real Linux kernel, then defend the evidence live.

Linux system status MCP, read-only, no mutation

Gymnasium environment where AI agents solve SPaRC grid puzzles one move per step, with optional backtracking. Comes as a pip package, a Prime Intellect verifiers env and an OpenEnv server. The best model solves 16.0%, humans 98.0%.

OpenTelemetry-native agent observability: trace every LLM call, tool call and agent step, with cost per call, alerts, and Claude Code / Codex / Cursor session tracing. Python SDK.

基于本地 Ollama 大模型的全离线 CTF 自动解题 Agent：单文件、纯 Python、四重安全闸门

随行 Suixing —— 手机端 AI Agent（Android · Kotlin + Compose）：读屏、点按、改文件、跑命令，纯本地、无云同步

Open-source enterprise AI agent: LLM tool-calling engine, permissions, RAG and a Python analytics engine, with a working ERPNext connector. Architecture guide: noviz.in/how-to-build-enterprise-ai-agent.html

📱 本地端侧手机 AI Agent：屏幕控件感知、多步意图操作、长期记忆沉淀，Android + 鸿蒙双端原生实现。

An AI security agent that really attacks: local audits + red-vs-blue drills in an isolated Docker range. 100% local, MIT. 一个会自己动手攻击的 AI 安全智能体：本机隐私/漏洞体检评分 + ReAct 智能体逐步带你处置 + 真实红蓝对抗演练场（Docker 隔离网内 Kali 攻击机 × WAF × 靶机，SQL注入真实报文）。全程本地运行，不出网，MIT 开源。

Official implementation of paper "Robotic Ultra-Long-Horizon Manipulation Skills via Human-guided Lifelong Code Generation"

An autonomous Lean 4 theorem-proving agent: LLM-drafted proofs iteratively repaired against lake compiler diagnostics with LSP goal-state feedback until type-checked. 68/100 on a graded 100-problem benchmark using a free 27B endpoint, $0.

Offline Linux command assistant: a 1.5B model on your machine (About 1 GB disk, 1 GB RAM, no GPU, ~2 s per answer) that checks every command against the man pages and tldr. No cloud, no API key. Rust + llama.cpp

Pi's coding agent ported to Rust

Agent skill: turn a short idea into a complete Krea 2 anime prompt with faithful Chinese translation.

MCP-сервер: LLM работает на Авито через отдельный Chrome — поиск, объявления, кабинет и Pro, профили, правка форм; подтверждение необратимых действий

基于 AgentScope 2.0 从零实现的企业级多智能体差旅助手

飞书知识库 RAG Agent：增量同步 · 父子切块混合检索 · JWT 多用户隔离 · 微调图表专家（FastAPI + LangGraph + Milvus）