About AgentMug

AgentMug is an AI worker platform by Dahshan Labs. Build a worker for a recurring task, connect the tools and information it needs, and review its results. Use the web app or work with the portable .agent file and developer tools.

What is an AI worker?

A worker is a reusable job definition: instructions, tools, and the accounts or knowledge it needs. It uses an AI model and available tools to carry out a task. Developer documentation also calls these workers AI agents. The actions a worker can take depend on its runtime, permissions, and configured connections.

Choose where it runs

The worker definition is portable. Model access, tools, and operating requirements depend on where you run it.

Choose where it runs
EnvironmentHow it works
AgentMug CloudBuild and manage workers in the web app. Hosted runs use managed model access and the cloud's supported connections, schedules, and run history.
AgentMug DesktopRun workers through the desktop app with your configured model access and supported local tools. Local execution depends on your computer and its setup.
AgentMug CLIInspect and run .agent files from the command line. Local runs use your model credentials; commands that call hosted workers need access to the host.

Connecting a service does not enable every action that service offers. Check the destination's tools and requirements, test a representative input, and review the output before relying on recurring runs.

What moves with a .agent file?

The .agent file is AgentMug's published JSON format for a worker definition. It describes the job and its requirements. Each owner configures their own accounts and sources; exporting a definition does not transfer cloud connections or make every tool available on another runtime. Keep credentials outside the file and review its contents before sharing.

.agent format and specification

Built by Dahshan Labs

Dahshan Labs makes AgentMug. AgentMug Core publishes the format, runtime, CLI, bridges, and reference examples. The hosted product and its managed services are separate from that public core. The repository explains the scope and licenses of each component.

Explore AgentMug