create-agent

To create a Harness AI agent, use an interactive workflow that gathers requirements and generates a standalone YAML specification with an agent.step.run block. This template includes container image definitions and inputs for llmConnector and anthropicModel.

A Harness AI agent specification is a standalone YAML template serving as a building block within pipelines. It enables automated code workflows, agentic tasks, and pipeline governance by defining reusable automation components.

You bind connectors to a Harness AI agent by defining optional MCP connectors and rules within the agent task block. The configuration workflow handles this binding during the interactive YAML specification generation process.

Yes, agent templates generated as standalone YAML specs can be deployed across multiple Harness pipelines. This enables consistent automation tasks, governance enforcement, and autonomous workflows throughout your software engineering projects.

Yes, the Harness AI agent YAML spec supports custom inputs. The specification requires inputs for llmConnector and anthropicModel, and allows optional MCP connector configurations to define the agent's runtime behavior and tool access.

The best way to update an existing Harness AI agent is through the interactive configuration workflow. It facilitates requirements gathering and YAML review to modify the existing agent.step.run block, inputs, and connector bindings.