AWS Bedrock

Connect an agent deployed as an AWS Bedrock AgentCore harness to AI GO! as a native model, using the typed aws_bedrock connection. The integration signs requests with your AWS credentials, mints a runtime session on the first turn, and round-trips the session_id automatically, so multi-turn evaluations preserve state.

Prerequisites

An agent deployed as an AWS Bedrock AgentCore harness, its region, and the harness ARN.
AWS credentials (access key ID and secret access key, plus an optional session token) authorized to invoke the harness.

Configuration

Define the model with connection_type: aws_bedrock and reference the AWS credentials as secrets:

key: "aws-bedrock-agent"
display_name: "AWS Bedrock AgentCore"
description: "Agent deployed as an AWS Bedrock AgentCore harness, via the typed `aws_bedrock` connection config."
task: "chat_completion"
rate_limit: 15
config:
  connection_type: "aws_bedrock"
  region: $AWS_REGION
  harness_arn: $AWS_HARNESS_ARN
  access_key_id: "<< secrets.AWS_ACCESS_KEY_ID >>"
  secret_access_key: "<< secrets.AWS_SECRET_ACCESS_KEY >>"
secrets:
  AWS_ACCESS_KEY_ID: $AWS_ACCESS_KEY_ID
  AWS_SECRET_ACCESS_KEY: $AWS_SECRET_ACCESS_KEY

Config fields

Field	Required	Secret	Description
`connection_type`	yes	no	Must be `aws_bedrock`.
`region`	yes	no	AWS region where the AgentCore harness is deployed.
`harness_arn`	yes	no	ARN of the AgentCore harness to invoke.
`access_key_id`	yes	yes	AWS access key ID.
`secret_access_key`	yes	yes	AWS secret access key.
`session_token`	no	yes	Optional AWS session token (for temporary credentials).

Credentials

The secrets block uploads values (from your .env or environment) to server-side secret storage, and << secrets.NAME >> references them from the config so no secret is stored in plaintext. Provide the values in your .env:

AWS_REGION=us-east-1
AWS_HARNESS_ARN=arn:aws:bedrock-agentcore:us-east-1:123456789012:harness/my-harness
AWS_ACCESS_KEY_ID=...
AWS_SECRET_ACCESS_KEY=...

For temporary credentials, also add session_token to both the config and secrets blocks.

Add and test

lf add model -f aws_bedrock.yaml
lf test model aws-bedrock-agent

Use in an evaluation

Reference the model key from a task specification in your run config:

evaluation:
  task_specifications:
    - task_key: my-task
      model_key: "aws-bedrock-agent"

Conversation state

On the first turn the integration mints a runtime session id; the session_id is attached to the assistant message and reused on each subsequent turn, so the agent remembers earlier turns.