AWS Bedrock models in Big-AGI

Running AWS Bedrock in Big-AGI

Connect Amazon Bedrock with your own AWS credentials and use frontier models inside the cloud your company already trusts. Big-AGI adds no markup and no intermediary: the billing relationship runs directly between you and AWS. Traffic keeps flowing between your setup and your AWS account, so your IAM policies, regions, and compliance story keep applying.

• Two ways to connect. The only integration here that isn't paste-an-API-key: Big-AGI supports both AWS's newer Bedrock API keys (sent as bearer tokens) and classic IAM credential trios with real SigV4 signing.
• Your account, your billing. Usage lands on your AWS bill at Bedrock's standard rates.
• One catalog, three AWS APIs merged. Big-AGI combines Bedrock's Foundation Models, Inference Profiles, and OpenAI-compatible listings into one de-duplicated catalog, so a model on just one API still shows up, and each source degrades gracefully.
• Claude on AWS. Run Anthropic's models through Bedrock alongside Llama, Mistral, Nova, and the rest of the catalog.

Why Big-AGI instead of the console?

The Bedrock console is built for one model, one prompt, one AWS account at a time. Big-AGI adds a real workspace on top of it: persistent chats, personas, and attachments, still routed through your own AWS credentials. It's also the only place a Bedrock-hosted model runs in Beam next to GPT, Gemini, and other frontier labs your account may not have in Bedrock at all.

Your keys and your data

Turn on Direct Connection and the browser talks to AWS directly with your credentials, skipping the Big-AGI server, when your setup keeps them client-side and Bedrock allows it. Your credentials stay in your browser. Chats are stored locally first, and sync only if you turn it on. The AI Inspector shows the exact request, the token counts, and a cost estimate against your AWS bill.

AWS Bedrock in Beam

Run a Bedrock-hosted model, Claude, Llama, Nova, whichever your account has access to, in parallel with GPT, Gemini, and more. Fusions then combine, cross-check, and synthesize the parallel answers instead of just picking the best one. Parallel runs use more tokens than a single chat.