CreateHarnessRequest
Types
Properties
The tools that the agent is allowed to use. Supports glob patterns such as * for all tools, @builtin for all built-in tools, or @serverName/toolName for specific MCP server tools.
Represents inbound authorization configuration options used to authenticate incoming requests.
A unique, case-sensitive identifier to ensure idempotency of the request.
The compute environment configuration for the harness, including network and lifecycle settings.
The environment artifact for the harness, such as a custom container image containing additional dependencies.
Environment variables to set in the harness runtime environment.
The ARN of the IAM role that the harness assumes when running. This role must have permissions for the services the agent needs to access, such as Amazon Bedrock for model invocation.
The name of the harness. Must start with a letter and contain only alphanumeric characters and underscores.
The maximum number of iterations the agent loop can execute per invocation.
The AgentCore Memory configuration for persisting conversation context across sessions.
The model configuration for the harness. Supports Amazon Bedrock, OpenAI, and Google Gemini model providers.
The skills available to the agent. Skills are bundles of files that the agent can pull into its context on demand.
The system prompt that defines the agent's behavior and instructions.
The maximum duration in seconds for the agent loop execution per invocation.
The tools available to the agent, such as remote MCP servers, AgentCore Gateway, AgentCore Browser, Code Interpreter, or inline functions.
The truncation configuration for managing conversation context when it exceeds model limits.