Skip to main content

AWS Bedrock Integration

Enterprise Feature

AWS Bedrock integration is available in AxonFlow Enterprise Edition. Full setup documentation including HIPAA-compliant configuration is available in the Enterprise Documentation Portal.

Contact Sales | Request Demo

Overview​

AWS Bedrock provides serverless access to foundation models from leading AI providers through a unified API. AxonFlow integrates seamlessly with Bedrock, enabling enterprises to leverage multiple LLM providers without managing infrastructure.

Why Use Bedrock with AxonFlow?​

  • Multi-Model Access: Claude 3.5 Sonnet, Llama 3, Mistral, and more through a single integration
  • Serverless: No model hosting infrastructure required
  • AWS Native: IAM-based security, CloudWatch monitoring, VPC endpoints
  • Cost Effective: Pay-per-token pricing with no minimum commitments
  • Compliance Ready: AWS compliance certifications (HIPAA, SOC 2, GDPR)

Architecture​

User Request
  |
AxonFlow Agent (Governance + Policy Enforcement)
  |
AWS Bedrock API (Model Routing)
  |
Foundation Models (Claude, Llama, Mistral, etc.)

AxonFlow adds governance and policy enforcement with minimal latency overhead (<10ms P95) while routing requests to Bedrock models.

Supported Models​

ModelModel IDUse CaseCost (Per 1M Tokens)
Claude 3.5 Sonnetanthropic.claude-sonnet-4-20250514-v1:0Complex reasoning, code generationInput: $3.00, Output: $15.00
Claude 3 Haikuanthropic.claude-haiku-4-5-20251001-v1:0Fast responses, high throughputInput: $0.25, Output: $1.25
Claude 3 Opusanthropic.claude-opus-4-20250514-v1:0Highest accuracy, complex tasksInput: $15.00, Output: $75.00

Recommendation: Start with Claude 3.5 Sonnet for balanced performance/cost, switch to Haiku for high-volume use cases.

Meta Llama Models​

ModelModel IDUse CaseCost (Per 1M Tokens)
Llama 3 70Bmeta.llama3-70b-instruct-v1:0Open-source alternative, codingInput: $0.99, Output: $0.99
Llama 3 8Bmeta.llama3-8b-instruct-v1:0Budget-friendly, simple tasksInput: $0.30, Output: $0.30

Recommendation: Llama 3 70B for cost-sensitive deployments where Claude's advanced reasoning isn't required.

Mistral AI Models​

ModelModel IDUse CaseCost (Per 1M Tokens)
Mistral Largemistral.mistral-large-2402-v1:0European compliance, multilingualInput: $8.00, Output: $24.00
Mixtral 8x7Bmistral.mixtral-8x7b-instruct-v0:1Efficient mixture-of-expertsInput: $0.45, Output: $0.70

Recommendation: Mistral for EU data residency requirements or multilingual use cases.

Prerequisites​

Before integrating Bedrock with AxonFlow Enterprise, ensure you have:

AWS Account Requirements​

  • AWS Account with Bedrock enabled
  • IAM permissions to enable model access and create policies
  • VPC with private subnets (required for HIPAA deployments)

Supported Regions​

  • us-east-1 (N. Virginia)
  • us-west-2 (Oregon)
  • eu-central-1 (Frankfurt)
  • ap-southeast-1 (Singapore)

Model Access Approval​

Bedrock models require explicit access approval (one-time setup):

  1. Open AWS Console -> Bedrock -> Model access
  2. Click "Modify model access"
  3. Select models you want to enable (Claude 3.5 Sonnet, Claude 3 Haiku, Llama 3 70B)
  4. Submit access request
  5. Wait for approval (instant for most models, 1-2 business days for some)

Deployment Options​

Deploy AxonFlow Enterprise via AWS Marketplace CloudFormation, which automatically:

  • Creates IAM roles with Bedrock permissions
  • Configures ECS tasks with appropriate execution roles
  • Sets up CloudWatch logging and monitoring

See AWS Marketplace Deployment for details.

Self-Hosted Deployment​

For self-hosted deployments, you need to create an IAM policy, attach it to your execution role, and configure AxonFlow environment variables.

AWS Credentials Setup​

Step 1: Create an IAM policy for Bedrock access

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "AxonFlowBedrockInvoke",
      "Effect": "Allow",
      "Action": [
        "bedrock:InvokeModel",
        "bedrock:InvokeModelWithResponseStream"
      ],
      "Resource": [
        "arn:aws:bedrock:*::foundation-model/anthropic.claude-*",
        "arn:aws:bedrock:*::foundation-model/meta.llama3-*",
        "arn:aws:bedrock:*::foundation-model/mistral.*"
      ]
    }
  ]
}

Save this as bedrock-policy.json, then create and attach the policy:

# Create the IAM policy
aws iam create-policy \
  --policy-name AxonFlowBedrockAccess \
  --policy-document file://bedrock-policy.json

# Attach to your ECS task role (replace with your role name)
aws iam attach-role-policy \
  --role-name your-ecs-task-role \
  --policy-arn arn:aws:iam::YOUR_ACCOUNT:policy/AxonFlowBedrockAccess

Step 2: Create an IAM role with a trust policy for ECS tasks

If you do not already have an ECS task role:

# Create trust policy
cat > trust-policy.json <<EOF
{
  "Version": "2012-10-17",
  "Statement": [{
    "Effect": "Allow",
    "Principal": { "Service": "ecs-tasks.amazonaws.com" },
    "Action": "sts:AssumeRole"
  }]
}
EOF

# Create the role
aws iam create-role \
  --role-name AxonFlowBedrockTaskRole \
  --assume-role-policy-document file://trust-policy.json

# Attach the Bedrock policy
aws iam attach-role-policy \
  --role-name AxonFlowBedrockTaskRole \
  --policy-arn arn:aws:iam::YOUR_ACCOUNT:policy/AxonFlowBedrockAccess

Step 3: Configure AxonFlow environment variables

Set these on the Orchestrator task definition:

AXONFLOW_LLM_PROVIDER=bedrock
AXONFLOW_LLM_MODEL=anthropic.claude-sonnet-4-20250514-v1:0
AXONFLOW_LLM_REGION=us-east-1
AXONFLOW_LLM_MAX_TOKENS=4096

Step 4: Verify access

# Confirm the role can invoke Bedrock
aws bedrock invoke-model \
  --model-id anthropic.claude-haiku-4-5-20251001-v1:0 \
  --content-type application/json \
  --body '{"anthropic_version":"bedrock-2023-05-31","max_tokens":10,"messages":[{"role":"user","content":"Hi"}]}' \
  /dev/stdout | jq .

Full setup instructions including HIPAA-compliant configuration are available in the Enterprise Documentation Portal.

Enterprise Features​

The following features require AxonFlow Enterprise:

HIPAA-Compliant Configuration​

  • VPC endpoint setup for private Bedrock access (traffic never leaves AWS network)
  • KMS encryption for CloudWatch logs
  • CloudTrail audit logging
  • Compliance checklist and verification

Multi-Model Routing​

  • Intelligent model routing based on query complexity
  • Cost-optimized routing policies
  • Fallback configuration for high availability

Cost Management​

  • Budget policies with daily/monthly limits
  • Cost tracking per tenant/agent
  • Automatic request denial when budget exceeded

Monitoring & Observability​

  • CloudWatch metrics integration
  • Prometheus metrics export
  • Cost and latency dashboards

Cost Comparison​

Bedrock models are typically 50-90% cheaper than OpenAI equivalents:

ScenarioOpenAI GPT-4Bedrock Claude 3.5 SonnetBedrock Claude 3 Haiku
1M tokens/month$40$18 (55% savings)$1.50 (96% savings)
10M tokens/month$400$180 (55% savings)$15 (96% savings)

Cost Considerations​

  • No minimum commitment: Bedrock uses pay-per-token pricing with no upfront cost or reserved capacity required.
  • Data transfer: Requests to Bedrock within the same region incur no data transfer charges. Cross-region calls add standard AWS data transfer fees.
  • VPC endpoints: Using a Bedrock VPC endpoint eliminates NAT Gateway data processing charges, which can be significant at high volumes (saves ~$0.045/GB).
  • Budget controls: AxonFlow Enterprise includes budget policies that can set daily or monthly token limits per tenant to prevent cost overruns. Community users should monitor usage via CloudWatch metrics.

Getting Started​

Ready to use AWS Bedrock with AxonFlow?

  1. Community Users: AxonFlow Community supports OpenAI and local LLM providers. Get started with Community

  2. Enterprise Trial: Contact sales for a 14-day enterprise trial with full Bedrock support. Request Trial

  3. AWS Marketplace: Subscribe to AxonFlow Enterprise on AWS Marketplace for immediate deployment.

FAQ​

Which Bedrock model should I use?​

  • Complex reasoning / code generation: Claude 3.5 Sonnet
  • High volume / cost-sensitive: Claude 3 Haiku or Llama 3 70B
  • Highest accuracy (worth premium): Claude 3 Opus
  • EU compliance / multilingual: Mistral Large

Does AxonFlow add latency to Bedrock calls?​

AxonFlow adds <10ms overhead (P95) for governance and policy enforcement.

Is Bedrock HIPAA compliant?​

Yes, with AWS BAA signed and VPC endpoints configured. Full HIPAA setup guide available in Enterprise documentation.

Can I use multiple Bedrock models simultaneously?​

Yes, AxonFlow Enterprise supports multi-model routing with automatic selection based on query complexity, cost constraints, or custom rules.

Last Updated: December 11, 2025

Last updated on