Skip to main content
Newton agents are preconfigured for specific Physical AI tasks, such as detecting anomalies in sensor data or checking a recording against a standard operating procedure. The agent blueprint defines what the agent does: the nodes, models, and settings that turn input into output. The section Canonical Blueprints below lists the agents we support. To build your own agent from the platform’s nodes, see Custom Agent; for help designing one, contact Sales.

How Agents Work

To work with Newton agents, you move through four phases. Each phase produces one object with its own ID, and each ID links back to the phase before it (by referencing that phase’s corresponding object). A session result can always be traced to the exact blueprint version that produced it.

The Agent Lifecycle

Diagram showing the workflow for using Newton
Agents

Creating and Using Agents

There are two ways you can create, monitor, and manage agents from these agent blueprints:

Canonical Blueprints

Agents are created based on agent blueprints, which are templates used as the starting point for building agents that use the Newton Platform to analyze sensor data. Archetype AI offers a variety of canonical blueprints you can use to create your own agents:

Next Steps

To create an agent with the Agents API, build a package with Create Package, deploy it with Create Deployment, then open a session on the deployment with Create Session.
  • Quickstart: run your first agent: copy a canonical blueprint and run a batch session on a sample file.
  • Tune a blueprint on your data: Optimize, evaluate, then promote the best trial to a new blueprint version.
  • Custom Agent: write your own blueprint from the platform’s agent nodes when no canonical blueprint fits.
  • Batch vs. serving sessions: run an agent either as a batch job that processes files and finishes, or as a serving agent that stays running to answer queries as you submit them.