paperclipai/paperclip
Node.js server and React dashboard that runs teams of AI agents as a virtual company, connecting them to outside tools through governed remote MCP connectors.
Awesome MCP › Agent Orchestration Platforms
invisible_dots is an open-source, self-hosted alternative to proprietary autonomous-agent products such as OpenAI Dots, Meta Muse, Grok Bot, Manus Cue, and Claude Cowork. Each agent instance, called a Dot, gets a hardware-accelerated virtual machine with a persistent Linux desktop: the VM sleeps when idle and wakes to handle the next message, task, or scheduled automation. A Dot's browser is a patched Firefox build (invisible_playwright_mcp, built on invisible_playwright and invisible_core) with the fingerprint altered inside the engine itself, aimed at evading anti-bot detection; each named browser identity keeps its own cookies and logins. Like Claude Code, a Dot writes its own notes and how-to skills into its home folder and reads them back on later tasks, and it can use any MCP server the user declares, whether a local program or a remote URL, with the server's keys kept out of the stored configuration — this is the project's MCP integration point. Every tool a Dot can call is assigned one of three permission levels (allow, ask, deny); an ask request waits in the user's inbox, survives restarts, and executes exactly once when approved. Work is organized as a priority queue with scheduled start times and recurring automations, and a crash or forced kill never loses an accepted message or re-runs a tool call. Users reach their Dots through a web UI, a command-line interface, an HTTP API with live events, Telegram, or an opt-in WhatsApp channel. The stack requires Node 24+, Go 1.25+, Git, and hardware virtualization, plus an OpenRouter API key for the underlying model calls; setup installs QEMU and the appropriate accelerator (KVM or Windows Hypervisor Platform) and downloads or builds the Dot VM images. The project is written in TypeScript and Go, licensed under MIT, and marked alpha status, with Linux and Windows as the supported hosts.
https://github.com/feder-cr/invisible_dots
Node.js server and React dashboard that runs teams of AI agents as a virtual company, connecting them to outside tools through governed remote MCP connectors.
LobeHub is a platform that enables multi-agent collaboration and human-agent co-evolution by treating AI agents as units of work, providing tools to build, manage, and collaborate with AI agent teams seamlessly.
Agent meta-harness that wraps Claude Code and Codex with 100+ specialized agents, coordinated swarms, persistent cross-session memory, and an MCP server so any MCP client can drive it.
OpenHuman is a local-first personal AI assistant with persistent Markdown memory, multi-agent workflow orchestration, and research tools, connecting to 100+ OAuth integrations and thousands of MCP servers.
GPT Researcher is an autonomous AI agent that conducts deep local and web research on any topic, generating detailed, unbiased reports with citations and supporting integration with the Model Context Protocol (MCP) for enhanced data sourcing.
Haystack is an open-source AI orchestration framework for building production-ready LLM applications in Python, enabling modular pipelines and agent workflows with explicit control over retrieval, ...
Activepieces is an open-source AI automation platform providing an extensible MCP toolkit and over 280 MCP servers for AI agents, designed for both technical and non-technical users.
Swarm is an educational framework by OpenAI for lightweight, scalable multi-agent orchestration using agents and handoffs, designed to explore and teach multi-agent coordination techniques.