Custom agents

Awesome list of projects for and with Raft (https://raft.build), the workspace where humans and AI agents collaborate.

harnsy turns Claude Code, Codex and OpenCode into a development team. You talk to one lead agent; it hands out tasks to the roles it needs, and brings back questions and finished work.

A Power BI specialist LLM for Ollama — local-first, grounded, 12/12 on a 12-task Power BI benchmark at ~156 tok/s. Built on gemma4:26b with a baked-in Power BI rulebook (DAX, M, PBIR, TMDL, safety gates) plus a unified RAG + gated-tools wrapper. PolyForm Noncommercial.

Book2Advisor — Turn a person's books/speeches/interviews into a traceable, inferential methodology advisor (Person Method Model + Method Transfer)

Tiny, zero-dependency, never-panicking HTML parser: CSS selectors, HTML→Markdown/text/JSON, reader mode and RAG chunks. Rust crate, htmlq-compatible CLI with a linter, npm (41 KB WASM) and PyPI packages.

Self-Healing Academic Knowledge Assistant powered by Agentic RAG, LangGraph, and Gemini.

The AgenC agent-to-agent coordination surface over MCP — browse the on-chain task marketplace, query the agent registry (ERC-8004-style identity), resolve three.ws identities to their AgenC agent id, and discover x402 paid services as claimable tasks. Read-first over the live public three.ws bridge. No key, no signer, no payment.

Cross-border brand UGC viral content synthesizer & ROI optimizer

Stop flooding your agent's prompt with MCP tools: MCPtoCodeMode puts every MCP server behind 3 tools and a sandbox, for Codex, Claude Code, Cursor, Grok, OpenCode, Pi, or any MCP client.

Recover broken Arabic from native PDFs with evidence-based diagnosis, conservative repair, BiDi/order recovery, artifact cleanup, and multi-column RTL layout. Zero-dependency core. OCR not claimed.

📘 基于视觉大模型与 RAG 的智能错题本：AI 拍照录题 · 结构化解析 · 向量检索举一反三 · SM-2 间隔重复复习 | Vision-LLM powered wrong-question notebook with RAG & spaced repetition

A curated, neutral list of tools, methods, and resources for building a second brain: personal knowledge management, note-taking, and personal AI.

麦门述职报告：用真实麦当劳订单生成年度绩效回顾 + 年度颁奖 + 省钱参谋

A tmux workspace tree and AI-agent monitor in one sidebar, with no daemon. Tracks Claude Code, Codex, OpenCode, Gemini, Cursor and more.

Open-core data platform with self-healing web scrapers — AI agents diagnose and patch broken extractors automatically (ClickHouse, Dagster, Qdrant)

⚡ Enterprise Spec-Driven Development (SDD) framework & native AI skill for converting codebases across stacks. Powered by 16 Universal Pillars for Antigravity, Claude Code, Cursor, Windsurf, Cline, Copilot.

An LLM-native programming language: you declare intent, and it lowers to a semantic IR that runs on an interpreter or compiles through MLIR to a native binary.

Pattern checks a coding agent's UI decisions against your own design system before it builds, whether that lives in code or in Figma. It helps the agent reuse what the product already has and build only what is actually missing. An opt-in enforcement boundary can make that check required, and every decision is recorded in an auditable ledger.

Token 超支？快给史密斯打电话！ Better Call Smith. Adaptive model routing, a glassy Codex usage monitor, and reproducible benchmarks with receipts.

Task queue and orchestrator for AI coding agents. Manage Claude Code agents from Discord — auto-recovers from rate limits, runs overnight, queues work across projects.

A ledger you text: WhatsApp → hledger, by text, voice or receipt photo

First-3-second shortform video hook retention scorer (Submagic / Opus Clip)

Скиллы для разработки в 1C:EDT по стандартам 1С (платформа 8.3)

GeoTask — 面向 AI Agent 的时空任务上下文引擎 | Spatiotemporal Task Context Engine for AI agents.

🪄 Install, review, and share agent skills.

Run Claude Code or Codex next to your real secrets: the AI sees placeholders, you see the real values.

An autonomous AI agent acting as a complete virtual comic studio. Powered by LangGraph and local Stable Diffusion 1.5 to generate short comic books from a single idea.

Agent orchestration for the Qwen Code CLI - 12 specialized subagents, 8 skills, Context7 + Grep MCP built in.

The runtime coat around agent loops.

Spec-Driven Development workflow system optimized for smaller models, with modular skills, clearer context separation, and human-in-the-loop task extraction.

⚡ Drop-in SKILL.md, .cursorrules, CLAUDE.md, AGENTS.md & Remote MCP (.mcp.json) pack for Claude Code, Cursor, Windsurf, Cline & x402 AI Agents — Verify GitHub Bounties & Mine DOOM/wDOOM on Base Mainnet.

⚡ The open ledger rules of the AAO Stack: append-only, hash-chained, integer minor units, idempotent writes, balances always derived. Flashy Gold is the first asset on the books. Apache-2.0 — npm install @flashylabs/ledger. The rules are open; the networks that run them live are products.

The complete coding agent for your terminal. Autopilot, decision system for screening and reranking, built-in token savings, background subagents/shell calls, learner + memory system, nudges that keep models on track (small and flagship), code mode, and everything else a modern coding agent should have, designed together in one native binary.

Flint: a local-first agentic workspace, a fork of Jan

GenOffice fork with a provider-agnostic AI slot — bring your own local LLM or OpenRouter key

AI memory system with an entity/memory graph, tool-calling agent, and connectors (GitHub, Slack, Linear, Notion, Gmail, Discord, Stripe...) — built in Rust on SurrealDB

【停止更新，新应用“ManJuX”】本地 AI 漫剧生产工作台 · Go 单服务集成小说管理 + 小说/脚本→漫剧全自动流水线（MiniMax H3 R2V）+ 智能体审片返工 + ComfyUI 自包含部署

Grok Bot OS: an open-source OS for Super Intelligence, where Grok Bot controls the whole machine

AI-assisted Logisim workbench for computer organization courses: draw, simulate, and let an AI agent read, wire, debug and explain your circuits. Compatible with Logisim 2.7.1 course packages.

AI coworkers built to get things done. They access your apps, operate your tools like a human teammate, and deliver finished results.

电气工程师 AI Copilot — 本地运行的电气工程师智能体工作台：AI 编写/读懂梯形图、直出多品牌 PLC 工程文件（信捷 .xdp / 三菱 / 汇川 / 西门子 / 松下 / 台达）、电气图设计与理解、智能选型、梯形图仿真器

Goal-driven mode for AI coding agents. Reach the user's win condition in the fewest moves.

An Orca best practice built around Matt Pocock's engineering skills. His chain stays: grill-with-docs, to-spec, to-tickets, then implement or implement-spec, closing with code-review. Orca supervises each step in its own worktree, with gates and merge. Not affiliated with or endorsed by Matt Pocock.

Vision-first desktop computer use for DSH on macOS: OCR text targeting, set-of-mark indices, 11 desktop_* tools and a visible cursor. No accessibility tree. · 让 DSH 看着屏幕操作你的 Mac。

Give your coding agent subagents in your other agent CLIs and accounts, all sharing the same tools. Plugin for Claude Code, Codex and Antigravity; MCP server for any other client.

Free Astro theme for AI agent and devtool startups. Monospace, light + dark, blog. MIT.

The AI assistant you just talk to (사사리사) — free, open-source voice assistant for macOS. Name it, say it: chores done on your Mac, coding handed to the AI you use (Claude Code, Codex, Aside, API or local models).

RepoLens parses public JavaScript and TypeScript repositories and turns their real imports into an interactive dependency map.

Tongji Online · 同济校园助手：以 Skill 分发，用自然语言查询 Canvas、图书馆、教务、一卡通与校园服务。

Tell your AI coding agent: plz do it, then prove it

Stop burning your Claude or Codex limits on boilerplate. Delegate multi-file implementation to Cursor's Composer 2.5 over MCP — your frontier model writes the brief and reviews the diff; Composer does the typing, fast, on a separate quota.

Orchestrate tasks execution using your coding agent of choice.

🪁 Generate with AI in the Panel, or connect your agents over MCP

A neutral contract and Go reference implementation for durable agent sessions: replay a run byte-for-byte with no model call, fork it for free, and verify the record independently.

A curated list of projects, workflow packs, examples, and integrations built with Smithers.

three.ws live discovery surface from any AI agent — trending agents and coins, the $THREE holder leaderboard with tiers, and the site-wide activity ticker. Read-only over the live public API. No key, no signer, no payment.

The command-line toolkit for Robinhood Chain — instant reads with zero config, guarded writes with a wallet. The gh/vercel of Robinhood Chain. npm: hood-cli.

Your AI memory bank is eating your context window. Capped, git-committed, retrieval-first project memory shared by every agent via AGENTS.md.

Go REST API scaffold: spec-first OpenAPI, chi, pgx/sqlc, OpenTelemetry, JWKS auth, embedded migrations, one-command resource generation.

Godot MCP server for AI-assisted game development｜连接 Codex、Cursor 等 AI 助手，运行游戏、检查场景与运行时状态、调试 GDScript，通过断言、日志和截图验证修改。

Judgment On Low-latency Thresholds. A under-100M-parameter reflex layer that decides agent steps in <10ms and escalates to an LLM only when uncertain.

Vitruvius is an engineering research agent.

QEMU harness for AI agents: boot aarch64/arm/riscv64/x86_64 kernels in seconds, exec via guest agent, fast rootfs rebuilds, IoT emulation

A production-oriented Levantine Text-to-Speech pipeline built on a fine-tuned XTTS-v2 optimized for real-time conversational agents.

AI Agent's Omamori — protect your system from dangerous commands executed via AI CLI tools

Computational Chemistry Meets Agentic Machine Learning

A customizable agent runtime system for external applications, with LangGraph-based agents, application integrations, Harness and Eval.

MCP ASCII Charts tool

GenPark AI Agent Skill - Conversational audio acoustic filler injector masking LLM inference and TTS latency with contextual conversational cues.

Runnable example guides for building AI agents and applications with Daytona sandboxes.

Untiring, token-efficient research collaborator: reads papers and patents, cites the paragraph, writes the paper, designs and simulates what it proposes.

Product Marketing skills for Claude Code and AI agents: agentic skills, commands, and plugins for Claude Code — from positioning to Go-To-Market strategy, growth experimentation, CI, and market research. Build one brain, every skill reads it.

The operating system built for AI agents. Purpose-built, open-source, engineered from the kernel up to run AI agents at hardware speed.

One browser shared by a human and an AI: visual for the human, source-level for the AI

Agent Muster brings Codex, Claude Code, Pi, Gemini CLI, OpenCode, Cursor, and more into one desktop or local-web workspace. Manage projects and sessions with multi-pane layouts, file/Git/terminal side panels, image and file references, and automate tasks via DingTalk, Feishu, WeCom, and WeChat bots.

The Cockpit for The Sovereign AI Software Factory

Reverse engineering, rebuilt for AI agents. A pure Rust engine your agent drives over MCP.

🎬 Type an idea, get a finished video. Claude researches the topic, writes and fact-checks the script, narrates with a cloned voice, builds real-footage, 3D or cartoon visuals, adds karaoke subtitles and music, makes the thumbnail and checks every cut. Free & open source.

A self-building personal agent with a Rust kernel and WebAssembly plugins. Light, free, roams across your machines, always on.

A desktop GUI for your coding agents. Run Claude Code, Codex, Cursor, OpenCode and more side by side, using the subscriptions you already have.

Local-first guardrails evaluation server for LLM apps. Scans inputs, outputs, and tool calls against a YAML policy and returns allow / block / redact. No database, runs entirely on your machine.

A terminal UI for browsing coding sessions across multiple tools from local transcripts, session logs, and databases. It currently supports Claude Code, Codex, and OpenCode.

Skills that turn AI code reviews into decision reports. You review the review, then decide.

Reference implementation and canonical explainer for ERC-8056 (uiMultiplier()), the corporate-actions standard behind Robinhood's Stock Tokens. npm: erc8056.

Click an element on your local dev site, write what should change, and your coding agent gets it — selector, computed styles, component chain, source hint and a cropped screenshot. MCP bridge for Claude Code, Cursor and Codex.

NEURONwriter MCP Server - Access SEO content analysis programmatically. Get NEURONwriter lifetime deals at softreviewed.com

A drop-in payment modal for any x402 paid endpoint. One <script> tag turns an HTTP 402 Payment Required into a polished checkout: discover the challenge, connect a wallet (Phantom on Solana, MetaMask / any EVM wallet via EIP-3009), sign, settle, and show a receipt — in vanilla JS, with no bundler and no framework.

Stripe-Checkout-grade USDG payments for Robinhood Chain — embeddable widget, hosted payment links, merchant verification library. npm: hood-pay.

Multi-tenant AI agents on AWS Lambda MicroVMs — one Firecracker VM per tenant with near-zero idle cost, snapshot resume, auto lifecycle, and EFS-persisted state.

Vellis — open-source context graph engine for AI agents: typed, local-first memory with explicit schema, validated change, deterministic query, migration, snapshots, replay, and audit. MCP-native. Apache-2.0.

A self-hosted AI factory with continuous context

AI-native game VFX: effects authored by agents over MCP, a local studio with a GPU viewer, a deterministic engine, and an MIT-licensed effect library.

Local-first HTTP API and MCP server that lets AI agents read and control TP-Link Tapo smart-home devices

面向公众号、小红书、抖音等自媒体封面及信息图生成的 Skill，推荐使用 GPT Image 2！

x402 payment facilitator — verifies and settles EIP-3009 micropayments multichain. v2 facilitator.

Claude Code status in the MacBook notch: live Clawd, hover details, Allow/Deny buttons

Communications infrastructure your AI agents operate. Skills plus Bird's hosted MCP server.

XClaw 是全球首个基于 语义拓扑（Semantic Topology） 的动态 AI Agent 网络基础设施。它为 Agentic Web 时代提供公共网络层，将全球分布的 AI Agent 节点连接成一个可发现、可路由、可协作的智能网络。