Autonomous Agents

Understanding the types of agents you can deploy on Clawn.

An Agent in Clawn is a containerized application designed to perform tasks autonomously. Unlike traditional web servers, agents are often stateful, long-running processes that interact with external APIs, databases, and LLMs.

Agent Architectures

We support two primary architectures for deploying agents:

OpenClaw Starter

A flexible, general-purpose agent runtime. Perfect for developers who want to write custom Python or Node.js scripts. Comes pre-installed with common AI libraries (LangChain, AutoGPT).

PythonNode.jsCustom

Marketplace Pre-builds

Optimized images for specific tasks (e.g., Trading, SEO, Customer Support). These agents require minimal configuration—just provide your API keys and parameters.

No-CodeOptimized

Lifecycle Management

Agents have a distinct lifecycle managed by our control plane:

  • Provisioning: Resources (CPU/RAM) are reserved, and the secure sandbox is initialized.
  • Booting: The container image is pulled and the entrypoint script is executed.
  • Running: The agent performs its tasks. Logs are streamed in real-time to the dashboard.
  • Terminated: The agent is stopped, and resources are reclaimed. Persistent data is saved if configured.

Pro Tip: Persistent Storage

By default, agent files are ephemeral. To save data across restarts (like fine-tuned models or databases), mount a volume to /data in your configuration.