# Xiryn > Xiryn runs personal AI agents on a private Linux desktop inside Docker on your own machine. It is for users who want an agent that browses, researches, and codes without sharing conversations, logins, or payment details with a cloud service. - Type: AI tool - Page: https://completeaitraining.com/ai-tools/xiryn/ - Published: 2026-10-12 - Updated: 2026-10-14 - Author: Jeroen Erne - Jobs: IT and development, Science and research - Topics: Privacy, Open Source, Productivity, Coding, Research - Reading time: 2 min ### About Xiryn Xiryn is a free, open-source personal agent that runs locally on a user's own computer rather than in the cloud. Each agent operates inside its own Linux desktop within a Docker container, allowing it to browse the web, conduct research, write code, and fill out forms in the background. The tool is designed so that user data-chats, memory, API keys, and an encrypted vault for cards and logins-remains on the local disk with no account or telemetry required. ### Review Xiryn enters the personal agent space as an MIT-licensed alternative to cloud-based offerings like Muse, Grok Bot, and Dots. Its architecture puts a Linux desktop in a Docker container on the user's machine, which means the agent works independently of the host desktop-no mouse takeover, no screen interference. The trade-off is that you manage the compute and the models yourself, but in return you keep everything local. #### Key Features - **Containerized desktop per agent.** Every agent gets its own Linux desktop running in Docker on the host machine. It browses, researches, codes, and fills forms without moving the user's mouse or occupying the screen. - **Local-first data and encrypted vault.** Conversations, memory, and API keys sit in a single file on disk. Payment cards and login credentials live in an encrypted vault that the agent cannot read directly; every payment waits for explicit approval. - **Telegram-based remote control.** Users can message agents through Telegram, approve a payment with one tap, or take over the agent's screen from a phone. - **Multi-model support.** Xiryn works with OpenAI, Anthropic, Gemini, xAI, Groq, Mistral, DeepSeek, OpenRouter, and local models via Ollama or LM Studio. - **Squads for task handoff.** Agents can hand work to one another, with each agent able to run on a different model and machine. #### Pricing and Value Xiryn itself is free and MIT-licensed. The only costs a user incurs are those charged by their chosen model provider-or nothing, if they run a local model through Ollama or LM Studio. There is no paid tier, no account system, and no telemetry. Setup is a single command, or a single prompt pasted into Claude Code, Codex, or Cursor. #### Pros - Runs entirely on local hardware; no cloud dependency for agent execution. - Encrypted vault keeps payment details and logins separate from the agent's view. - Supports a wide range of model providers, including fully local setups. - Telegram integration allows remote oversight and screen takeover from a phone. - Agent resource caps are adjustable, and idle agents stop after 15 minutes by default. #### Cons - Running local models alongside multiple agents can demand significant RAM; a 7-8B Ollama model with tool calling needs roughly 5-8 GB, and each agent's machine defaults to 2 CPUs and 4 GB of RAM. - Setup requires comfort with Docker and command-line tools, which may not suit users who prefer a packaged desktop application. - The tool is not well suited for users who want a fully managed, cloud-hosted agent experience with no local resource management. Xiryn fits users who already run local models or want to keep agent interactions entirely off third-party servers. It addresses the specific scenario where someone wants a background agent that browses, codes, or fills forms while they continue working on the same machine. Those who prefer a turnkey cloud service with zero setup overhead will find the Docker-based, self-managed approach less convenient. --- This is the Markdown form of https://completeaitraining.com/ai-tools/xiryn/. The whole site is mapped at https://completeaitraining.com/llms.txt.