Connect AWS Bedrock with an AWS credential

Bedrock takes an AWS credential and a region, issued by your own AWS account. Generate a long-term key - the short-term kind expires with the console session and cannot list models, which is what leaves the list empty on an otherwise correct setup.

The AWS account behind it

Usage lands on the AWS invoice. A funded AWS account is the prerequisite for a first answer. The enforceable limit is an AWS Budget on that account.

Generating a long-term key

1. Open the Bedrock console in the region you intend to use. Switch regions first - the credential is scoped to one.
2. API keys -> Long-term API keys -> Generate long-term API keys -> pick an expiration -> Generate (AWS's own steps).
3. The IAM route instead: in IAM, a user with the AmazonBedrockFullAccess policy -> Security credentials -> Create access key -> Application running outside AWS.

The two credential modes

Credential mode	Fields to fill
Bearer token	Bedrock long-term API Key, plus the region
IAM	AWS Access Key ID and AWS Secret Access Key (session token optional), plus the region

The modes are mutually exclusive and the app enforces that: a bearer token wins, and the IAM fields are dropped from the request. The region defaults to us-east-1.

To fill them, open Models (Ctrl + Shift + M), press More Services if the Setup AI Models wizard opens, then Add -> AWS Bedrock. Complete one mode plus the region and press Models (add and manage your keys).

Region and model access decide the list

Each model must be enabled for your account in that region before it can appear. Together with the region scope, those two conditions account for nearly every empty list here.

Every request takes the server path

This API sends no browser-access headers, so every request goes through the Big-AGI fast edge servers.

Related

• Connect an AI service
• Keys held by a server you operate
• Direct Connection