Custom agents

Provide a lightweight agent kernel designed for efficient, fast startup in distributed AI applications with minimal resource use.

🤖 Build intelligent, offline LLM agents with LangGraph and llama-cpp-python using this starter template for local, private tool-calling applications.

RAG-based internal knowledge copilot for fintech support, providing grounded answers with source citations and safe refusals.

Secure local AI agent runtime — vault-backed secrets, sandboxed execution, cryptographic audit chain, procedural skill system

🤖 Build personalized intelligent agents easily with AgentX, a platform that simplifies creation using natural language and integrated tools.

Drive Claude Code on your Mac from your phone: list and resume every session, start new tasks, stream results live.

Feed LLMs financial data and interactive charts in one call — an MCP server for A-share quotes, financials, and macro indicators, 42 tools with results-as-UI.

Talk to your music library. A Rust CLI that turns natural-language queries into Spotify and YouTube Music actions — explore your catalogue, build playlists conversationally, and migrate between providers, all driven by an LLM agent with signed, scoped permissions.

🎨 Enhance cinematic image quality with ComfyUI-None-upup. This AI engine offers nodes for clarity, brightness, and video processing to elevate your visuals.

MCP server for n8n workflow management -- view, create, sync, and manage n8n workflows via AI assistants

Agentic framework for quantitative finance & algorithmic trading

A skill to make agents search for better models

Multi-agent orchestrator that runs a hierarchical engineering team (Product → EM → Tech Lead → specialists) on top of locally-installed Claude or Codex CLIs.

Production AI code review + autonomous web research agent. Claude tool-use loop over 4 Bright Data products (Web Unlocker · SERP API · Scraping Browser · MCP Server). HubSpot CRM hub, Lead Brief in Telegram with freshness buckets. Oracle Cloud free tier, $0/month, solo founder.

Record and preview video from USB-C capture cards directly in the macOS menu bar using hardware-encoded HEVC.

The persistent home a stateless AI agent can't host for itself — register once, resume your whole self in one call; real browser; cited deep research; hire real humans for physical-world tasks; free keyless chat + ID portraits. MCP + REST, x402 on Base, free to start.

Voice-based coding assistant speak commands, get AI work + narrated feedback in real-time

UAHP v0.5.4: Solving the "Silent Agent" Problem with Death Certificates and Liveness Proofs

Ecosystem-aware CLI for validating and upgrading Agent Skills: enforce spec compliance, detect vendor extensions without breaking, convert large skills into progressive disclosure.

This project is a GitHub Repository Analyzer that helps developers quickly understand unfamiliar repositories. Often when developers encounter a new repository, it can take significant time to understand the structure, technologies used, and architecture. So I built a web application that analyzes a GitHub repository and generates insights

The ultimate open-source, local-first AI agent for job search. Find, score, tailor, and track career opportunities using a multi-agent self-correcting QualityCritic loop.

llmpress reduces the token cost of sending source code to an LLM by detecting repeated patterns, replacing them with short aliases, and transparently restoring the originals in the model's response.

Rust SDK for MnemeBrain — a belief memory system for AI agents. Supports MnemeBrain Lite and Full APIs.

Portable AI-workflow control plane for Claude Code — one install.sh ships specialized agents, user-invocable skills, always-on rules, lint/audit/secret hooks, the 6-gate review, design templates + sprint discipline, an /onboard wizard, and a pull-based feedback loop. v0.4.0

Stop non-deterministic edits: A strict "Read-Match-Hash-Write" pipeline for safe and reproducible code modifications. AnchorScope is a reference implementation of a deterministic code editing protocol for LLM agents.

Computer Use for Windows - without the screenshots. Element-based MCP server: UIA + Win32 + CDP, 3-tier fallback ladder, ~ms per command.

Sandbox any CLI AI agent so it runs unrestricted without touching your machine. Written in Rust, daemonless, no Docker: each box is a disposable git worktree plus an in-process libcontainer OCI container. Spin up several agents in parallel against one worktree. Built for Claude Code, Aider, Gemini CLI and friends.

Everything you need to build AI agents and AI-powered products with Mollie.

An open-source AI agent that builds the tools it's missing, mid-task — under your governance. Pull capabilities at runtime instead of pre-wiring them. Local-first, MIT.

H-SPAE AlphaAgent: session-persistent harness for LLM-driven alpha factor mining with memory, resume, context compression, and reproducible artifacts.

Persistent local MCP/CLI memory system for AI agents. CPU-only, 98.94–99.15% R@5 on LongMemEval. Named after Mnemosyne (Greek: μνήμη, memory).

Runtime control and observability for AI agent actions: policy, approvals, durable boundary evidence, outcome assurance, idempotency, and OpenTelemetry correlation.

Asmbli

Public architecture and product documentation for Anion — a local AI operating layer for the Linux desktop.

Engineering architecture for runtime state modulation in Large Language Models.

Local-first LLM operating-system hub for BACH, Rinnsal, gardener, SQLite agent memory, MCP servers, skills, and multi-agent orchestration

Define rules over a read-only data source; Silmari safely derives review-priority signals (실마리) for a human to decide — read-only · scoped · audited · local-first.

Trajectory-level observability and evaluation for LLM agents. Trace, score, and detect regressions in multi-step tool-using agent runs.

A Docker-based isolation harness for running AI coding agents

Notes and small projects from running multi-model LLM agents: deployment, model routing, and cost control.

Epistemic reasoning engine for AI Agents — assess claim credibility with calibrated confidence. Includes a ready-to-use .skill package for agent frameworks. 87%+ accuracy with evidence, 85%+ abstain without.

Self-hosted NL analytics API: query planner, conversational Q&A, Vega-Lite charts, guardrails, and optional vector RAG with Python & TypeScript SDKs.

Java game development framework in Minecraft style

🧠 MindPeek | 让 AI 真正读懂人心 告别传统用户画像方法！ MindPeek 通过 LLM 深度解析，从每轮对话中自动提取性格、偏好、潜在需求——构建持续进化的个人画像。 ✨ 核心能力 • 多维度特征提取：性格特点 / 价值观 / 潜在想法 / 情感状态等 • 多 Agent 协作架构：自我进化，越用越懂你 懂用户的 AI，从这里开始。

Agent-first document ingestion layer on Docling with Markdown, image sidecars, manifests, OCR, and quality gating.

LangGraph-native product-analytics copilot: plain-English questions in, auditable task DAGs out, executed in parallel against an embedded DuckDB warehouse

LLM-agent-native backtest framework for futures research — MCP server, in-package skills, catalogued error codes, typed Pydantic configs. Production engine inside Qorka @ DolphinQuant.

Find the layer where a language model commits a decision — and steer it. Any open-weight HF model. (WANDERING arc paper #6)

Prepare job applications with a team of AI agents that do the legwork while you make every call. Self-hosted, bring your own AI account. The AI drives, while you hold the wheel.

一个面向 AI Agent / 量化实习的多智能体量化研究平台 / A resume-ready multi-agent quantitative research platform for AI Agent & Quant internships

Expand your LLM and AI agent capabilities with this collection of modular tools and specialized skills.

Multi-Agent Code Evolution Platform — debate-style review with health scoring, tech debt tracking, architecture analysis, and improvement roadmaps

An open-source AI desktop companion with a living 3D character, desktop awareness, voice interaction, memory, and intelligent automation.

Generate production-grade technical specifications from feature requests using AI coding agents.

Arcone-Agent: Rust-native agent toolkit for DeepSeek, Python, RAG, tools, sessions, and streaming.

Monorepo for Glyph, EI semantic capsules, bridge evals, and unified-loop workflows for auditable agentic control.

An AI-native KMS and next-gen RAG prototype. Enforces strict tree-shaped index constraints and compiler-like diagnostics after every write, enabling agents to self-heal knowledge structures and eliminate garbage-in, garbage-out RAG pain points.

Free, public event-sourced world model & knowledge graph for multi-LLM agents — propose/evaluate/commit visions to shared state, query a live causal knowledge graph via REST + MCP (11 tools). Live at insidedcpulse.com

稳健型个人加密货币量化系统。freqtrade 地基 + 自建薄壳，纪律写进状态机，LLM 只做研究/复盘/否决。

🏥 Clinical Decision Support System for Pulmonary Embolism - An agent-based AI system using LLMs and RAG to assist healthcare professionals with evidence-based PE diagnosis and treatment recommendations.

Companion page for A Survey of Harness Engineering: the ETCLOVG seven-layer taxonomy of agent harness engineering, with pointers to a public catalog of open-source projects.

An open-source, LLM-agent-powered social media simulation platform for research

Explainable Anomaly Detection for ICS - SWaT Testbed

Polymathera's no-RAG, multi-agent framework for extremely long, dense contexts (1B+ tokens).

Run multiple AI agents simultaneously in separate terminals with built-in chat on Windows, macOS, and Linux using TypeScript and Electron.

AI-powered multi-agent PM-KISAN eligibility assessment system | BTech AIML PBL-2

Do agents leave footprints? A system for deliberate orientation when minds enter long-lived territories.

Anvil-code — autonomous coding agent that runs entirely on your machine. 32B local models + structural-anchor edits, no cloud.

Agent context compressor and memory persistent orchestration

A living narrative world engine — AI showrunner, multi-character scenes, memory, evolving world state. 一个会记忆、会演化的 AI 剧本杀世界。

Adversarial AI system to test and improve reliability under real-world pressure

Terminal-based multiagent coding assistant for C, C++, Python, and reverse-engineering

Roteador Inteligente de CLI Agents — Nunca mais escolha qual ferramenta usar

A beginner-friendly OpenClaw multi-agent setup wizard with safe Feishu onboarding, guided mode selection, and automatic starter profile generation.

AI-powered boardroom for founders. Describe your startup idea — 40+ specialist AI agents analyse it across strategy, GTM, finance, legal, and risk simultaneously. Agents disagree with each other, and a Chief of Staff resolves the conflicts into a health score, blind spots, and ranked next steps.

Trilha aberta e gratuita, do zero ao avançado, para construir assistentes educacionais inteligentes: NLP, Machine Learning, Deep Learning, Transformers, LLMs, RAG, agentes e Learning Analytics.

Remote-control interactive agent CLIs (Claude Code, Codex, Gemini) over tmux. One binary, three faces, one provider abstraction. Subscription-billed.

A study of production multi-agent patterns — planning, reflection, RAG, MCP, human-in-the-loop, guardrails, and evals — wired into one Streamlit orchestrator on the Claude Agent SDK.

OPEX (ОРЕХ) — самостоятельно развёртываемый AI-шлюз на Rust: каждый слой заменяем без изменения ядра — агенты в Markdown, инструменты в YAML, провайдер одной строкой. 150+ провайдеров, автоокна для 5000+ моделей, мульти-агентная оркестрация, каналы, память на pgvector, шифрованный vault. Без vendor-lock.

Autonomous DFIR agent driving SANS SIFT forensic tools via an MCP server + LangGraph, with guardrails and a hash-chained audit trail.

多 Agent 群聊系统 — AI 协作、@mention 链、LangGraph 工作流

Autonomous multi-agent trading system in Go with LLM-powered analysis, debate, and execution pipelines

Plugin-first conductor for multi-agent code work. Compile a plan into slices, run each in an isolated worktree, fall through Claude → Codex → Gemini.

A local-first runtime for running many distinct agents — each with its own memory, secrets, skills, workspace, and autonomy — from one install.

A Feishu-first multi-agent system for content intake, evaluation, and action routing.

Multi-agent AI healthcare-navigation assistant for Indian families — find care, check schemes, read documents, set reminders, and book appointments, safely. Built on Google Cloud & Gemini.

An ADR-driven security framework for Kali — a HackTheBox co-pilot orchestrating Hat-persona Claude agents across a shared arcade and a cross-engagement memory layer.

产品开发团队模拟技能 - 7+N 角色虚拟团队 + Phase-Gate 质量门禁 + 标准/资源受限/MVP 三种运行模式

Coordinate multiple Claude Code terminals on one project — file locks, shared kanban, planner approval gate. Battle-tested in production.

Self-hosted autonomous AI agent platform with 90+ MCP tools, persistent memory, and self-extension capabilities. Runs on your remote machine and grows its own capabilities.

Repository of Scripts Related with SBGames paper

Connect AI agents representing humans through the A2A protocol for seamless, secure interaction across apps and tools.

Claude Code marketplace and home of claude-engineer: an opinionated, design-system-first, plan-gated autonomous engineer for full greenfield React/Next and FastAPI projects.

A research agent that runs for hours. Question + deadline in, cited Markdown report out. Single machine, local LLM.

Persistent autonomous goals for Pi — with no-progress detection, evidence-based completion, token budgets, and auto-continuation

v0.15.7, 508 tests, 234 archived FIDs. Autonomous AI-powered DEX trading engine — Arbitrum via 0x API. Funnel v1 production stack (momentum pre-scorer + top-K), 10-LLM multi-model jury on NVIDIA NIM. Reasoning-first JSON with LONG/SHORT/NO_SIGNAL.

Nikola — autonomous AI system based on ATPM consciousness architecture. Nitpick is its primary language substrate.

Uniswap V4 Hook - loyal traders earn cheaper swap fees via on-chain karma score. Autonomous AI agent, X Layer Testnet, 75 tests passing.

An open-source coding agent for your terminal — hosted frontier models or fully offline via Ollama. One account with Aether AI.

🦔🫛 HedgePod is a ⚛️ 🏆 ETHGlobal Buenos Aires 🇦🇷 Prize Pool Winner 🌎 World Mini App that solves crypto's biggest UX problem: chain fragmentation. Autonomous cross-chain DeFi that makes 23M World App users their own hedge fund. Deposit once. AI agents automatically rebalance across 8+ chains for optimal yield. Gasless. Built for non-crypto use