# How to Configure Amazon Bedrock Models in OpenMAIC > Learn how to configure Amazon Bedrock models in OpenMAIC by setting environment variables or defining a server-providers.yml block. Easily integrate powerful AI models. - Repository: [MAIC/OpenMAIC](https://github.com/THU-MAIC/OpenMAIC) - Tags: how-to-guide - Published: 2026-09-08 --- **Set `BEDROCK_REGION` and `BEDROCK_MODELS` as environment variables (or define a `bedrock` block in [`server-providers.yml`](https://github.com/THU-MAIC/OpenMAIC/blob/main/server-providers.yml)) to enable the Bedrock provider; OpenMAIC then automatically activates it via `applyBedrockProviderConfig` in [`lib/server/provider-config.ts`](https://github.com/THU-MAIC/OpenMAIC/blob/main/lib/server/provider-config.ts).** Configuring **Amazon Bedrock models in OpenMAIC** requires understanding that the platform treats providers as **server-managed** services with explicit opt-in. Unlike other providers that may auto-register, Bedrock is only activated when you supply configuration through environment variables or YAML. This guide walks through the exact mechanism, configuration options, and runnable examples drawn directly from the THU-MAIC/OpenMAIC source code. ## Activation Mechanism: How Bedrock Gets Enabled OpenMAIC's provider system gates Bedrock behind a deliberate activation check. The function **`applyBedrockProviderConfig`** in [`lib/server/provider-config.ts`](https://github.com/THU-MAIC/OpenMAIC/blob/main/lib/server/provider-config.ts) (lines 69–94) implements this logic: ```typescript const yamlBedrock = yamlProviders?.[BEDROCK_PROVIDER_ID]; // line 69 const envRegion = process.env.BEDROCK_REGION?.trim() || undefined; // line 72 const envModels = splitModels(process.env.BEDROCK_MODELS); // line 73 const hasExplicitBedrockEnv = !!envRegion || !!envModels || …; // lines 74-79 ``` If **no configuration source** provides Bedrock details **and** `BEDROCK_REGION` is absent, the function returns early (lines 85–87), completely omitting the provider. This prevents accidental Bedrock usage. When activated, the merged configuration lands in `config.providers[BEDROCK_PROVIDER_ID]` (lines 89–94), which the LLM factory in **[`lib/ai/providers.ts`](https://github.com/THU-MAIC/OpenMAIC/blob/main/lib/ai/providers.ts)** consumes: ```typescript const bedrock = createAmazonBedrock({ region: resolveBedrockRegion(), credentialProvider: createBedrockCredentialProvider(), }); ``` ## Configuration Methods You can configure Bedrock models through **environment variables** (recommended for deployment) or **YAML configuration** (preferred for version-controlled setups). ### Environment Variable Configuration Set these variables in `.env.local` or your deployment environment: | Variable | Purpose | Required? | |----------|---------|-----------| | `BEDROCK_REGION` | AWS region (e.g., `us-east-1`) | **Yes** — enables provider | | `BEDROCK_MODELS` | Comma-separated model IDs | **Yes** — defines available models | | `BEDROCK_API_KEY` | Custom endpoint API key | Optional | | `BEDROCK_BASE_URL` | Non-AWS endpoint URL | Optional | | `AWS_BEARER_TOKEN_BEDROCK` | Bearer token credential | Optional | ```bash # .env.local BEDROCK_REGION=us-east-1 BEDROCK_MODELS=us.anthropic.claude-sonnet-5,us.anthropic.claude-opus-4-8 DEFAULT_MODEL=bedrock:us.anthropic.claude-sonnet-5 ``` The **`splitModels`** function parses `BEDROCK_MODELS` into an array. Environment variables take precedence over YAML values when both are present. ### YAML Configuration Alternative Create or edit [`server-providers.yml`](https://github.com/THU-MAIC/OpenMAIC/blob/main/server-providers.yml) for infrastructure-as-code setups: ```yaml # server-providers.yml providers: bedrock: models: - us.anthropic.claude-sonnet-5 - us.anthropic.claude-opus-4-8 # apiKey: your-key # optional # baseUrl: https://... # optional ``` Note that `BEDROCK_REGION` in environment variables still activates the provider—YAML alone without any env var won't trigger `hasExplicitBedrockEnv` to true unless region is also specified in YAML. ## Setting the Default Bedrock Model Client-Side After server-side configuration, users can select Bedrock models. Set `DEFAULT_MODEL` with the `bedrock:` prefix: ```bash DEFAULT_MODEL=bedrock:us.anthropic.claude-opus-4-8 ``` Or programmatically in client components: ```typescript import { useModelStore } from '@/stores/model'; useModelStore.setState({ modelId: 'bedrock:us.anthropic.claude-opus-4-8' }); ``` The model ID string follows the pattern `bedrock:` where `` matches an entry in your `BEDROCK_MODELS` list. ## AWS Credential Handling The **`createBedrockCredentialProvider`** function in [`lib/ai/providers.ts`](https://github.com/THU-MAIC/OpenMAIC/blob/main/lib/ai/providers.ts) resolves credentials through: 1. Standard AWS SDK credential chain (IAM roles, `~/.aws/credentials`, etc.) 2. Optional `AWS_BEARER_TOKEN_BEDROCK` environment variable for token-based auth No additional configuration is required if your server runs with standard AWS credentials. ## Complete Configuration Example Here's a production-ready `.env.local` for OpenMAIC with Amazon Bedrock: ```bash # Core Bedrock activation BEDROCK_REGION=us-west-2 BEDROCK_MODELS=us.anthropic.claude-3-5-sonnet-20241022-v2:0,us.anthropic.claude-3-opus-20240229-v1:0 # Default model for new conversations DEFAULT_MODEL=bedrock:us.anthropic.claude-3-5-sonnet-20241022-v2:0 # Optional: proxy configuration for VPC endpoint # BEDROCK_BASE_URL=https://vpce-xxx.bedrock-runtime.us-west-2.vpce.amazonaws.com # Optional: bearer token if not using standard AWS credentials # AWS_BEARER_TOKEN_BEARROCK=eyJhbGc... ``` ## Key Source Files Reference | File | Purpose | |------|---------| | [`lib/server/provider-config.ts`](https://github.com/THU-MAIC/OpenMAIC/blob/main/lib/server/provider-config.ts) | `applyBedrockProviderConfig` — merges env/YAML and conditionally activates Bedrock | | [`lib/ai/providers.ts`](https://github.com/THU-MAIC/OpenMAIC/blob/main/lib/ai/providers.ts) | `createAmazonBedrock` — instantiates the Bedrock client with resolved config | | `.env.example` | Template with Bedrock placeholders | | [`README.md`](https://github.com/THU-MAIC/OpenMAIC/blob/main/README.md) | AWS credential handling documentation | ## Summary - **Enable first**: Set `BEDROCK_REGION` to activate the provider via `applyBedrockProviderConfig` - **Define models**: Use `BEDROCK_MODELS` (comma-separated) or YAML `providers.bedrock.models` - **Precedence rules**: Environment variables override YAML values - **Client selection**: Prefix model IDs with `bedrock:` in `DEFAULT_MODEL` or `useModelStore` - **Credentials**: Standard AWS SDK chain or `AWS_BEARER_TOKEN_BEDROCK` for custom auth ## Frequently Asked Questions ### What happens if I only set BEDROCK_MODELS without BEDROCK_REGION? The Bedrock provider **will not activate**. The `hasExplicitBedrockEnv` check in [`lib/server/provider-config.ts`](https://github.com/THU-MAIC/OpenMAIC/blob/main/lib/server/provider-config.ts) (lines 74–79) requires `envRegion` to be truthy, or the function returns early at lines 85–87. You must set `BEDROCK_REGION` to enable Bedrock. ### Can I use Bedrock with a custom endpoint or proxy? Yes. Set `BEDROCK_BASE_URL` to your proxy URL and optionally `BEDROCK_API_KEY` for authentication. These merge into the provider configuration's `baseUrl` and `apiKey` fields, respectively, as handled in `applyBedrockProviderConfig`. ### How do I switch between multiple Bedrock models in the UI? Ensure all desired models are listed in `BEDROCK_MODELS`, then toggle via `useModelStore.setState({ modelId: 'bedrock:' })` or configure `DEFAULT_MODEL` for the initial selection. The UI populates available models from the server-managed provider list. ### Does OpenMAIC support AWS SSO or temporary credentials? Yes. The `createBedrockCredentialProvider` function uses the standard AWS SDK credential provider chain, which automatically handles SSO profiles, IAM roles, and temporary credentials from `~/.aws/credentials` or environment variables.