What is Xiryn?
Xiryn is a free, open-source personal AI agent that runs on a Mac, Windows, or Linux computer, with agents operating in separate Linux desktops inside Docker. It can browse websites, research, write and run code, fill forms, and handle scheduled routines, while users can control it through its browser app or Telegram. Founders, researchers, and others who want an agent on their own machine can choose from model providers or use a local model, and approve payments before they go through.
Xiryn Features:
Dedicated Agent Desktops: Each agent runs in its own Linux desktop in Docker, separate from the user's regular computer session, and its desktop can be reset without removing its memory.
Recurring Routines: Users can describe recurring jobs in plain language, set schedules and time zones, and receive completion notices on their desktop, in the app, or through Telegram.
Remote Telegram Control: Users can message agents through Telegram, approve payments with one tap, or take over an agent's desktop from a phone.
Model and Cost Controls: Xiryn supports eleven model providers, including local models, shows costs by reply and totals by day, model, and agent, and can pause paid models at a set budget.
Encrypted Credential Vault: Saved logins and payment cards are encrypted on disk with a key held by the operating system; credentials are masked from agents and filled only on the relevant or approved site.
Agent Squads and Memory: Agents can have separate roles, models, desktops, and long-term memory; squads can route requests to a suitable agent, and users can review and edit memory notes.
Xiryn Benefits:
Keep Your Workspace: An agent can handle browser tasks in its own desktop without taking over the user's cursor or interrupting work on the main computer.
Private Local Use: Using a local model keeps prompts on the computer and avoids model API charges, which suits people who want local processing without usage bills.
Manage Model Spend: Cost records and budget limits help users understand paid usage and stop it from continuing past a chosen amount.
Review Before Paying: Payment approval gives users a chance to check a proposed purchase before an agent completes the transaction.
Fewer Repeated Checks: Scheduled tasks can take routine monitoring off a user's daily checklist and deliver results when a run is complete.
Use Cases:
Morning News Brief: Set an agent to prepare a brief on selected topics each morning and send a notice when the routine finishes.
Weekday Job Search: Have an agent look for new roles on weekdays and tailor applications to suitable listings.
Trip Planning: Use an agent to compare flights and stays, then prepare a day-by-day travel plan.
Price Checks: Schedule an evening price check or ask an agent to buy an item after reviewing and approving the payment.
Lead Research: Assign an agent to research prospective customers and gather information for outreach.