Skip to main content

Overview

Terminal Use is a platform for hosting long-running AI agents that need access to a filesystem. It provides everything you need to go from prototype to production:
  • Python & TypeScript SDKs - High-level APIs for building agents
  • AgentServer - Simple decorator-based runtime for agent logic
  • CLI Tools - Deploy, manage, and monitor agents from the command line
  • Platform - Zero-ops hosting with versioning, secrets, and observability

Key Capabilities

Simple Agent API

Decorator-based handlers for create, event, and cancel lifecycle

Claude Integration

Deep integration with Anthropic’s Claude Agent SDK

Managed Deployment

Zero-ops hosting with Docker and Kubernetes

Features for Production

Versioning, rollbacks, secrets, and multi-environment support

Core Concepts

Agents

An Agent is a deployed unit of logic that can handle tasks. Agents are:
  • Containerized FastAPI servers (Docker images)
  • Versioned (branches and releases)
  • Scalable (multiple replicas)

Tasks

A Task is a single execution context for an agent. It’s akin to a conversation. Each task is associated with a particular version of an agent. If you update your agent, tasks can be migrated to the new version of your agent.

Filesystems

Filesystems provide persistent file storage that exists independently of tasks (think of them as folders on your computer). When creating a task, you configure which filesystems to mount thereby giving the agent access to specific folders for that task. This enables you to:
  • Share files across multiple tasks
  • Enable multiple users to access common assets (e.g., shared team resources)
  • Allow users to download files produced by the agent
  • Pre-load files for the agent to work with

Namespaces

Namespaces provide a physical isolation boundary for compute and storage. If you serve other businesses, each one of your customers can have their own namespace to ensure compute and data are fully isolated.

Next Steps