Bedrock Guardrails MCP Server

Description

This MCP server exposes all AWS Bedrock Guardrails management features as tools, allowing you to create, edit, delete, and export guardrails (including all policy types) programmatically or via LLMs like Claude Desktop. It also supports exporting guardrail configurations as Terraform files for infrastructure-as-code workflows.

Features

Create, update, delete, and list AWS Bedrock Guardrails
Full support for all policy types: content, topic, word, sensitive information, contextual grounding, etc.
Export any guardrail as a Terraform .tf file (CloudFormation schema)
Secure authentication using AWS credentials from environment variables
Ready for integration with Claude Desktop or any MCP-compatible client

Prerequisites

Python 3.12+
uv package manager
AWS account with Bedrock Guardrails permissions
AWS credentials set in your environment (see below)

Setup

Clone the repository

Install dependencies:

uv venv
source .venv/bin/activate
uv pip install -r requirements.txt  # or use pyproject.toml with uv

Set AWS credentials in your environment:

export AWS_ACCESS_KEY_ID=your-access-key
export AWS_SECRET_ACCESS_KEY=your-secret-key
export AWS_SESSION_TOKEN=your-session-token  # if using temporary credentials
export AWS_REGION=us-east-1  # or your preferred region

Security Note: Never commit your credentials to source control. Use environment variables or a secure credentials manager.

Run the server:
```
uv run server.py
```

Claude Desktop Integration

To use this server as a tool in Claude Desktop, add the following to your claude_desktop_config.json:

{
  "mcpServers": {
    "BedrockGuardrailsMCP": {
      "command": "uv",
      "args": [
        "--directory",
        "/Users/haji/mcp-servers/bedrock-guardrails-mcp",
        "run",
        "server.py"
      ]
    }
  }
}

Replace the directory path with your actual project location.
Restart Claude Desktop after updating the config.

Example Usage

Create a guardrail with full policy config: Use the create_guardrail_full tool, passing all required and optional policy configs as dictionaries.
Update a guardrail: Use update_guardrail_full with the guardrail ID and any fields to update.
Export to Terraform: Use export_guardrail_to_terraform(guardrail_id) to get a .tf resource block for your guardrail.

Screenshots

Impact & Use Cases

Enterprise LLM Safety: Centrally manage and automate Bedrock Guardrails for all your generative AI applications.
DevOps & IaC: Export guardrails to Terraform for version control, reproducibility, and CI/CD integration.
LLM Tooling: Enable LLMs (like Claude) to programmatically manage safety policies, audit, and remediate risks.

Security

Credentials are only read from environment variables and never stored in code or logs.
All AWS API calls are made using the official boto3 SDK.

License

See LICENSE file.

AWS Bedrock Guardrails