Custom agents

Desktop profile for the DeepSeek Harness: a native Tauri window over the harness web surface — tray, single-instance, OS notifications, suggest-only updater, native dialogs, drag & drop, and an in-app settings tab.

Open-source AI app idea generator. Answer a few questions and get detailed product specifications, tech stacks, and MVP plans for your next startup or side project.

DeepSeek Harness 桌面端 —— 自包含的 ElectroBun 外壳，内置可自动更新的 dsh 引擎，附带 Windows 图形化安装程序

Educational Vue/FastAPI demo for structured symptom intake and pluggable AI providers. Not medical advice.

HELIX — Hybrid Envelope-multiparticle LInac eXplorer: TraceWin-compatible linac simulation with 3-D PIC space charge, matching, a PyQt6 workbench, and a voice-driven AI copilot. Built at Fermilab for PIP-II.

Product thinking, wired into your coding agent. AI-native product development framework for Claude Code - 50 skills spanning discovery, validation, domain modeling, JIT feature delivery (FDD+SDD), a client discovery layer for commissioned builds, and a reconciliation-based Rebuild playbook for existing products.

Turn any AI coding agent into a KiCad hardware design assistant - instruction library + stdlib-only verification tools that catch what ERC misses

A secure, containerized Model Context Protocol (MCP) server that bridges any MCP-compatible client with a full-featured Kali Linux environment for authorized security auditing.

Token-frugal JVM source navigation for AI. Find classes & methods in Gradle/Maven jars via on-demand decompilation. MCP + CLI.

MCP server for Magpie Capital — gives LLMs and agents native tools for the Solana lending protocol: live stats, collateral catalog, loan terms, x402 agent endpoints.

A fast Rust tool for harvesting source files as Markdown and copying them to your clipboard.

Structural prompt analysis + output testing + rewriting. Model-agnostic via LiteLLM. FastAPI service.

AI-powered platform — fintech app builder (Claude/GPT-4o/Gemini streaming) + cozy image generator. Built with zero human code using skillforge-ai.

Organize your Mac Downloads folder by project, not by file type. Renames opaque PDFs (arXiv IDs, UUIDs) to their real titles. Claude Code skill + CLI. No delete in the codebase; one-command undo.

Touch ID confirmation for AI coding agents on macOS. Pair with sandboxes for defense-in-depth against prompt injection.

Restore and upscale images in ComfyUI using native nodes for the SeedVR2 1.4B model.

Salesforce AI Assistant — An intelligent, LLM-powered assistant that enables natural language interaction with Salesforce metadata and data. Built with LangGraph for agent workflows, Weaviate for semantic search, MongoDB for short-term memory, and high-quality embedding models. Supports SOQL generation, metadata exploration, and secure multi-org

Open-source biomedical research agent with 60+ offline tools: variant interpretation. Use via web chat UI or MCP (DeepSeek-harness compatible).

Self-owned TerminalBench 2.0 coding-agent harness with a Rust Worker, Harbor/verifier-grounded evaluation, and guarded Heuristic Learning.

Autonomous long-form serial fiction with auditable LLM agents, persistent narrative state, immutable revisions, and reader-mechanism research.

The agentic, local-first job-application agent — runs on Ollama, never submits, never lies.

A markdown-based agent operating system for AI-assisted cold outbound. Draft-first: nothing sends without explicit human approval. Point it at any market.

Voterpool is an open-source consensus layer for AI-agents

Local-first macOS app that indexes your files and builds an AI-organized virtual folder tree without moving or renaming originals.

Give coding Agents runtime evidence for debugging and verifying DSH / Cordis plugins. 让 Coding Agent 获得用于调试和验证 DSH / Cordis 插件的运行时证据。

Self-hosted single-user web app for operating a private second-brain vault with an AI executive assistant.

BeeBoard is an AI-assisted beekeeping management platform built with FastAPI and Vue, covering apiaries, hives, inspections, honey batches, sales, tasks, and Docker-based deployment.

devspace-AI prepares your Mac for development with GitHub Copilot/agent workflow

Integrate an AI assistant into a DevExpress BI Dashboard.

AI-powered terminal assistant that suggests or executes bash commands from natural language.

ΠΑΝ — Personal AI Network. Always-on voice assistant that records you and controls your devices. Open source. Self-hosted.

Brain-Computer Interface control system — routes neural signals to desktop, mobile, IoT, and embedded devices with AI-powered command execution

Context broker for AI coding agents (CLI + MCP). Queries codebases via progressive disclosure instead of feeding entire files into context — reducing token waste and context rot. AST-aware retrieval, surgical editing with syntax guard, semantic search. Inspired by Recursive Language Models research (MIT/Stanford, 2025). Built in Rust.

MATLAB-compatible runtime for humans & AIs

Durable Playwright + Chrome task runner for AI agents: one-file jobs, persistent Profiles, multi-agent isolation, background execution, and partial results.

Servidor MCP de facturación electrónica española con VeriFactu (AEAT): emite facturas F1/F2, correctivas R1–R5 y valida NIF desde agentes de IA (Claude, ChatGPT, Cursor). Spanish e-invoicing / VeriFactu MCP server.

Independent, open-source Novamira companion plugin — safe, purpose-built MCP abilities for WooCommerce: variable products, media library, orders, and RankMath SEO.

Assistente Virtual Localhosted

Nook is a progression system for AI agents. Verified work generates immutable sparks that evolve your agents over time

Event-based long-term memory for LLM agents — associative surfacing over query RAG. Closed-loop episodic events, budgeted injection, tiered forgetting. Claude Code hooks + DeepSeek Harness plugin, one shared .memory/ format.

Chain-of-density study of 'Better Harnesses, Smaller Models: Building 90% Cheaper Agents via Automated Harness Adaptation' (Yang, Zhao, Wu & Kästner 2026, CMU, arXiv:2607.08938) - five-tier note, locator-verified. SLM agents, meta-agent harness optimization.

Autonomous engineering control plane: a five-tier hierarchical multi-agent DAG (Supervisor > Managers > Workers > Validator > Reducer) that follows software changes from PR to production - evidence-driven, provenance-preserving, spec-driven.

Noetrium: Reproducible Research Infrastructure for AI Agents - experiments, models, environments, evidence, recovery, evaluation, and verification.

Official code for "Every Answer Has Its Own Path: Agentic Routing for Retrieval-Augmented Table Question Answering" (EMNLP 2026)

Multi-agentic biomedical literature evidence extraction, curation, and summarization from PubMed/PMC in Open Knowledge Format (OKF) for downstream machine learning, feature selection, and agentic applications.

Compile human language into token-efficient machine-readable task packets for LLM agents

Agent skills and runtime plugins maintained by Chris A. Kim

Software engineering agent workflow as an executable graph: 22 skill nodes, 33 edges, a CI-runnable validator, a resume-capable runner protocol, and 30 drop-in skills

🚀 ODIN Reflector 2026 – Next-Gen Meta-Reflection Engine for Outline-Driven AI Development

AGENTS.md / CLAUDE.md behavioral contract for coding agents, Claude Code, Cursor, Codex, and LLM software engineering. Turn skills into instincts.

Plan-mode alternative for OpenCode — hand off the context window, not a plan document. ~50% cost, ~95% of frontier pass rate.

Declarative test-case management: git-native case files, agent execution, human review gate

Autonomous coding harness with a hardened, sandboxed execution environment

OpenClaw 源码解析 — 现代 Agent 系统架构

Self-healing web crawler — compile your LLM scraper into free, deterministic code that auto-heals when a site's layout changes.

Governance layer for multiple AI coding agents on a production codebase: constitution, decision log, evidence grammar, CI gates.

Reference factory for role-based agent workflows with requirements discovery, architecture gates, observability, tests, and human approvals.

EiHAI builds

Acrylamide — native Rust Agent Development Environment

AI-powered LeetCode assistant Chrome extension for learning data structures, algorithms, and coding interviews.

ACP coding agents (Cursor and Claude) that live in Emacs Org buffers.

On-device AI utility assistant powered by Gemma + LiteRT-LM with Wikipedia summarization and QR code generation.

Gemini Roleplay Core systemprompt, a framework for highly immersive, psychologically profound Simulation\role-playing.

Self-RAG chatbot with LangGraph that retrieves from internal PDFs, verifies grounding (IsSUP), checks usefulness (IsUSE), and iteratively rewrites queries to improve answer quality.

OpenClaw low-memory optimization guide for resource-constrained servers (2GB RAM)

MCP server for reverse engineering — 77 semantic operations: static triage, unpacking, decompilation (Ghidra/IDA), dynamic instrumentation (frida/TTD), emulation (Unicorn), symbolic execution (angr), evidence handling. Works with any MCP-compatible agent host.

Stateless Model Context Protocol (MCP) server for Fastly Compute — an edge-native, WebAssembly reference implementation of the 2026-07-28 spec.

Let your agent drive your Android phone — from Windows, over ADB, with nothing installed on the phone. Works with Claude Code, Codex and Hermes.

Document-grounded adaptive tutoring with bounded tool-use agents

A reusable skill & plugin kit for long-running AI coding agents: file-backed delivery contracts, discriminative validation gates, and restart-safe evidence. Deploy paths for Claude Code, Codex, TraeCode, and DSH.

Evidence-gated vulnerability research for AI security agents.

AI agent炒股荐股盯盘+open claw微信实时推送播报工作流

《深入理解 AI Agent》工程方法论蒸馏 Skill：39 张能力卡，1 个入口，三重验证+盲测 48/48

A self-hosted job-hunting agent driven from Discord: finds postings, scores them, drafts cover letters, tailors your CV, and can watch your inbox -- runs on a local LLM.

empirical research framework designed to analyze multi-agent Large Language Model (LLM) conversations as observable, continuous-state dynamical systems.

Self-hosted deployment for PlanVault — governed AI agent execution runtime. Docker Compose stack, signed images, Cosign verification, single-tenant VPC deployment.

Message board for LLM agents on one machine: one C binary, a plain-text log, and a recv that wakes on inotify instead of polling. Zero-token idle. A blackboard, not a broker.

Local traffic runtime for Claude Code and Codex — capture, route, transform, and inspect AI provider traffic on macOS.

Local Pokemon Red LLM-agent harness with live gameplay UI, structured traces, save states, replay, and turn-by-turn agent observability.

Resource management planner

Digital Engine for the development of Social media and digital marketing strategies, social media plans,, their implementation, and the monitoring and evaluation of the same

Spec-Driven Development (SDD) for Claude Code

Evidence-gated Recursive Self-Improvement (RSI) runtime for self-improving LLM agents, self-evolving agents, agentic AI, AI safety, evaluation, lineage, and rollback.

A multi-tenant platform that provisions and routes isolated DeepSeek Harness (dsh) instances — one container per user.

Autonomous AI code review & PR automation — a self-hosted, multi-agent software factory that reviews, fixes, and merges its own pull requests through a verifying LLM pipeline.

Static security scanner and analyser for AI agents, MCP tools, and LLM applications - OWASP LLM Top 10, and infrastructure checks. Zero runtime dependencies.

Native desktop chat platform — Discord-like, but self-hosted on your own Supabase. Rust P2P voice/video, an AI assistant that runs your server, realtime translation, and a loot-box economy. Tauri 2 + React + Supabase.

A co-pilot for your first real circuit board. Reads your KiCad project, explains what the checks are telling you, and sizes a case that fits. Works alongside KiCad and FreeCAD.

Full-API MCP servers generated from official OpenAPI specs: Stripe (587 tools), Vercel (333), Clerk (224). One npx command, intent-friendly tool names. Run standalone or behind Toolport.

Turn Instagram reels into local transcripts, keyframes, and LLM-ready Markdown bundles.

Proton Pass credential provider for agent-browser — resolves logins so the password never enters the agent's transcript, argv, or shell history. Zero runtime dependencies.

Native C++ coding agent built around a small core, lazy-loaded capabilities

Workflow-net runtime for DeepSeek Harness: resource-aware concurrency, native loops and fan-out, static soundness checking before a plan becomes durable, and process mining over its own event log.

Runnable, offline reproductions of one defect class in three agent frameworks: a value that is neither an explicit yes nor an explicit no reaches the execute path of a human-approval gate.

Humans fly to space. AI does the work. — Autonomous coding agent orchestrator that makes small, committed, documented changes while you sleep

Grok Build Web UI

A lightweight agent framework for production runtime. Cycle-based execution with pluggable LLM backends, tool dispatch, memory compression, and distributed scheduling.

基于 Claude Code 架构精简的 AI Agent 框架 | 7层架构 Agentic Loop + 6个内置工具 + 三级权限控制 + 动态System Prompt | 可以更加详细的学习工程级架构,也可以自定义内置工具使用

Ticket-Driven Development for AI agents — the agile, lightweight memory + backlog framework that lets Claude always know what to work on next.

Learn OpenCode — From Zero to Professional. Interactive course with 6 modules, 17+ lessons, 5 hands-on projects. Agentic coding with MCP, rules, custom agents, skills, and free AI models. Built by Ayman Elmasry — AEL Digital Studio.