Custom agents

Open-source agentic AI for social media planning, scheduling, approval, and browser-based publishing.

A universal, cross-platform logging system for AI agents.

A Agentic AI-Powered Newsletter Generation System Built for the Portia AI Hackathon

WhatsApp-first accounts-receivable agent for invoice reminders, payment events, and human-reviewed collection workflows.

A capability lockfile for local models. Catch silent tool-calling regressions in CI (deterministic, no LLM judge).

🥧 Apple Pie — AI coding agent harness for native Android & iOS apps: drives Claude Code to take a Jira or markdown ticket through plan → implement → verify and out as a reviewed pull request. Android ships today; iOS is under development, preview coming soon. Parallel agents, one TUI dashboard, never auto-merges.

The Debian COSMIC Desktop Brewery: Unified .deb and systemd-sysext delivery for Debian Testing/Sid.

Open-source graph-native workbench for governed AI workflows — execution graphs, durable context graphs, and installable Industry Packs.

Future of Software

DSH-Desktop - DeepSeek Harness Web GUI 的 Windows 桌面版

Yana AI 🐰 One runtime. Any AI. Human-governed. A local-first, cross-platform system for running, connecting, orchestrating, and governing AI — with deterministic control over what it can access, change, and execute. Your AI can act. But who decides how far it can go?

Official implementation of SKILLER: Language-Level Reinforcement Learning for Reusable Skill Extraction in Small Language Models.

Do LLMs Know What to Ask and When? Evaluating Multi-Turn Information Seeking

HxSync: An Android AI Agent client with 4-tier scheduling (local/API/remote/gateway), MCP-compatible memory, and human-in-the-loop confirmations. Bring your own LLM.HxSync 是口袋里的 AI Agent 客户端：四档调度（本地/API/远端/网关），MCP 记忆内核，人在回路确认卡。自带模型，也可接你自己的 AI 服务。

A modernized fork of Stable Diffusion WebUI Forge — Gradio 6, Python 3.12, transformers 5, self-bootstrapping single-script deploy.

Spring Boot + Vue 3 教务管理系统 / academic management system for graduation projects, with Redis 抢课, course selection, grades, exams, timetable, AI assistant, natural language SQL, and load testing.

Maestro is an open-source, multi-agent AI development framework written in Go. It orchestrates specialized AI coding agents across a deterministic software engineering workflow

Yang Code Review (YCR) - A lightweight GitHub Action for AI-powered, OWASP-aligned code review with your own LLM provider.

Run models that do not fit in your RAM. Keeps the always-read weights resident, streams the routed experts from disk.

A curated directory of open-source AI repositories for agents, LLM apps, coding assistants, MCP servers, RAG, generative AI, automation, developer tools, research, and AI security.

面向《崩坏：星穹铁道》的配装配队优化：已成体系的战斗模拟 DSL 文档与多源数据管线；仿真引擎（干净设计：事件总线 + 闭合关键字 DSL）与 autoresearch 式 agent 开发中。

Empathetic AI engineering partner with cognitive memory, multi-agent orchestration (10 agents), 132-signal skill routing, and self-improving reflection loops. Built on Model Context Protocol. By Aldo Karendra.

Evidence-driven Kubernetes root-cause analysis agent built with LangGraph and MCP

Bounded Agents: Delegation Security for Multi-Agent AI Systems - Reference Implementation and Evaluation Artifact

Multi-agent orchestration for pi; chain, parallel, and dependency-graph delegation with per-step model routing, preset-driven tool scoping, and background execution with live monitoring.

Production reliability, benchmarking, and evaluation layer for MCP agent runtimes — protocol negotiation, A2A interop, circuit breakers, connection pooling, adaptive streaming, and autoscaling, with measured throughput/latency and a 109-test suite.

Nine-agent AI framework automating quant research from hypothesis generation through leakage-aware backtesting

Chat-first film preproduction skill for Codex: director-room ideation, story/script/shot design, visual review, model-specific prompts, and verified releases.

Complete reference guide for Microsoft Copilot Studio — Instructions, Knowledge, Tools, Agents, Topics, Activity, Evaluation, Channels, Analytics + IT Service Desk Bot POC

Structured multi-step orchestration for MCP agents: adversarial lenses, reusable SOP workflows, and persistent traces.

Lossless wire format for agent-to-agent handoffs. 36.6% fewer real cl100k_base tokens on held-out ToolBench, every byte of decode overhead counted, byte-exact round trip.

Multi-agent orchestration for DeepSeek Harness

Centralized role-based model, thinking, and tool routing for Pi Fabric

多 Agent 组织型全栈开发集群运行时（Python + LangGraph）：12 岗位 / 7 类会议 / 审批门 HITL / YAML 流程 DSL / 六步自我进化闭环

Shared HTTP control plane for Chronicle traces and TokenOps spend: one SQLite, one UI. Agents never open the database file.

deepseek harness 纯牛马 —— 多智能体数字员工指挥台插件

Bootstrap, maintain, and continuously improve a coding-agent workspace so each session adds momentum the next session inherits. For Claude Code + Codex CLI.

Local CLI and daemon for supervising collaboration between Claude Code, Codex, Grok, Gemini, and other AI coding agents over MCP. Supports work directory sandboxing preventing unexpected writes by reviewing agents.

A map showing where the pieces for growing an AI family live. Non-runtime, no install, every part works on its own.

智多星 · 企业级多 Agent 平台 — 自然语言驱动知识问答、数据分析与业务执行，RBAC + 审批 + Saga 护栏 | Enterprise multi-agent platform: RAG, data analysis & business execution with RBAC, approvals & Saga

Several AIs debate a topic while you moderate — self-hostable, and it runs on your own local models. Open-source (AGPL).

Multi-agent coordinator for herdr: a head Claude Code agent that delegates tasks to worker agents (claude, codex, gemini, ...) in their own tabs

one coordination board for your team's coding agents. claude code, cursor and codex share a job board and take a lock on a file before editing it.

Model-agnostic autonomous coding agent engine with Claude Code capabilities, subagents, and an 18-target cross-agent SSOT harness

LWA Conduit — DAG parallel orchestration with multi-CLI role routing. Repo: lwa-conduit · PyPI: kiro-conduit

Persistent, goal-driven UltraWork mode for pi.dev — auto-continues across turns and fans sub-tasks out to child pi processes.

One shared Markdown file that AI agents take turns writing to — context, decisions and progress all live in the file, so any agent can pick up the work without you summarising the last chat. Single binary: CLI + MCP server + live Web UI

Unity ML-Agents reinforcement learning experiment where an autonomous agent learns to chase a target and avoid obstacles.

A structured agent runtime for LLMs. Structured context, multi-stage workflows, and ECS-based orchestration in a single binary

A flexible, modularized and easy-to-use agent framework.

A from-scratch Python agent harness: tool orchestration, multi-agent routing, session lifecycle, event-driven hooks, and cross-session memory recall.

Agentic MCP server for European seismic hazard & risk. 24 tools over ESHM20+ESRM20: hazard, UHS, CMS, fragility, damage, AAL, conditional-spectrum record selection (ESM), user-defined GMMs per TRT, browser front-end. OpenQuake-backed, stateful 4-tier cache. Companion to Vemula, Jehel, Cotton, Gatti (2026).

An open-source, AI-powered application using Agentic CAG to chat with any public GitHub repository or developer profile, offering deep code analysis, visual architecture maps and security audits.

Evidence-tiered positions on AI-native software development. Every claim traces to a primary-verified source; every position states what would overturn it.

OAKit (Office Agent Kit) gives AI agents a reliable way to read, understand, edit, and generate PowerPoint, Excel, and Word documents through a unified structured model—making document analysis, transformation, and automation easier to build into agent workflows.

A hands-on GenAI Engineering roadmap from LLM fundamentals to production AI systems. Covers prompt engineering, RAG, Agentic AI, LLM evaluation, observability, AI security, LLMOps, production deployment, CI/CD, multimodal AI, MCP, cloud, and AI system design through practical projects and real-world architecture.

MEOK AI Labs MCP Server

25+ dependency-light Go packages for CLIs, terminal UIs, and agentic tools. Wrapped in one module, designed so AI coding agents write correct code the first time.

Local-first control plane for AI-agent development — watch every Claude Code session live, ledger tasks/worktrees/resources, track token spend and cost, enforce rules.

Self-hosted AI copilot for OpenWrt router monitoring, diagnostics, management and network operations.

Same you, every coding agent.

dsh·古法编程插件

Cross-cutting design & typography skill engine with an anti-AI-slop doctrine — deliberate, human-crafted type, colour, layout & UI. Referenced by every Chwezi engine.

Kaira: A multimodal AI agent built with React and Python. Features include local wake-word detection, system registry management, autonomous coding tools, and persistent memory

Livis（理想同学）与 Hermes Agent 的非官方社区适配器

IQ-free LLM-controller for online automatic modulation recognition (AMR): the LLM plans deterministic RF tools and never sees raw IQ. Code, results, and figures.

Benchmarking search agents for closed-corpus, multi-hop retrieval over visually rich documents.

Your genome, labs, prescriptions and wearables sit in silos that never compare notes, so contradictions go unnoticed. This reads them against each other, with the source behind every answer — one local engine behind a web UI, a CLI, a Claude skill, and an Ouroboros plugin.

Company Is a Word. 一句话开一家AI公司 - open-source multi-agent orchestration framework for non-developers. 小白5分钟拥有自己的AI公司，可DIY任意行业、调教子Agent、无限家公司，支持飞书遥控。MIT

👹 malwi - no more supply-chain attacks

MusePi — 桌面优先的 AI 编程助手:Electron GUI + TUI/CLI + daemon 服务 + 桌面伙伴。原生视频理解、MCP、advisor、记忆、七域技能、自有更新渠道。

Native Windows desktop client for the pi coding agent - AI-powered coding assistant with embedded browser, multi-tab sessions, image support, fork tree, and glass themes. Built with Electron + React.

Open-source, self-hostable social media management & marketing platform (Django + React) — scheduler, content calendar, analytics, unified inbox, click-to-WhatsApp bot builder, AI assistant. MIT.

Privacy-first AI meeting assistant. Captures mic + system audio, transcribes live, and writes the summary — entirely on your machine. Tauri/Rust core, transcribe.cpp for STT, bundled llama.cpp. No account, no cloud, no telemetry.

Bo cham diem AI Agent dung Skill/Tool cho AutoCAD - Civil 3D - Revit BIM cau. 48 case, 6 hard gate, 197 test cho chinh bo cham.

CyberAgent — agent orchestration & human-agent collaboration: agentic project and task management where coding agents and people ship work together, with review gates, shared context and full audit. By Hyperspace.

Self-hosted command center for orchestrating AI coding agents

Personal, self-hosted task orchestration across heterogeneous devices, agents, and hardware.

GitSquad: A multi-agent orchestration framework for autonomous software development workflows on GitHub.

A platform for studying how LLM agents evolve communication under constraint — multi-agent simulations, an event-sourced run log, and post-hoc evaluation metrics.

浏览器仪表盘：在 PTY 下拉起你本机真实的 claude / codex / opencode / reasonix / zulu 命令行，组成一个协作蜂群（共享收件箱 + 黑板 + 队长调度），并叠加多模型研究委员会与融合竞赛。Rust + React + Tauri。

Research-oriented cognitive AI architecture exploring memory, reasoning, planning, and adaptive learning using local LLMs.

Cursor Composer 2.5 powerpack for Pi: agent chains, dispatch teams, Pi-Pi experts, live subagents, Pi-to-Pi coms, damage-control safety, curated agents and mono-black theme. Install with pi install. Requires Pi ≥0.75.3.

Active Inference simulation framework for ant colonies and other decentralized multi-agent systems — models foraging, trail formation and task allocation with multiscale sensory/motor/learning agents, and maps runs onto real ant species for validation.

Git-native multi-agent orchestration: agents don't share thoughts, they share verified work.

Aurogene - self-evolving digital mind

Spec-driven agentic workflow for AI coding agents, with specialized subagents (research, plan, critique, execute, verify, debug) via markdown specs

A kernel for concurrent multi-agent software development. Allows multiple agents to edit a shared worktree simultaneously

🛰️ Autonomous multi-agent AI trading system for XAUUSD (Gold) & BTCUSD on MT5 — quant ML ensemble (XGBoost/LightGBM/LSTM/HMM) + GenAI bull/bear debate + PPO reinforcement learning + live behavioral risk manager. 9 microservices, Next.js dashboard.

A curated collection of WorkBuddy use cases

Agentic AI co-pilot for crypto-exchange financial crime investigations: entity-network mapping, RFI contradiction checks (roadmap), and regulator-grounded SAR drafting over a tamper-evident audit trail. Synthetic-data research prototype.

MAS-AI Technologies website for governed AI products, research, and enterprise services.

Agentic AI for biologically grounded morphological feature design from microscopy — designs, implements, and validates compact cell-profiling features. Includes a desktop Qt UI demo.

Daimon: The Personal AI Conscience for the Agentic Era — an open-source framework for building a conscience layer for personal AI agents

Rust SDK for ENSIP-25: Link verification between ENS names and ERC-8004 AI agent registries.

Local, private job-search board. Runs in Docker, no setup. Add jobs from the board, the command line, or by handing a job link to an AI assistant that checks eligibility before it files anything.

DSH Desktop —— DeepSeek Harness 的 Windows 桌面客户端 + 模型路由 / mmx 图像理解插件（绿色免安装 · 自动更新）