AI Code Assistant Solutions

Amazon Bedrock has been selected as the organization's approved AI model provider for the EU region. Amazon Bedrock delivers access to a diverse selection of integrated LLM models with strong data privacy and regional compliance guarantees.

Multiple AI code assistant tools are compatible with Amazon Bedrock, including OpenCode, Continue.dev, and Claude Code, allowing developers to choose the tool that best fits their workflow while maintaining organizational compliance standards.

This solution was selected based on several key advantages:

• EU Data Residency: EU inference profiles ensure data stays within European boundaries
• Model Flexibility: Access to various large language models including Claude, Llama, and others
• Tool Compatibility: Works with multiple IDE extensions and desktop applications
• Security & Compliance: Enterprise-grade security with AWS infrastructure
• Robust Feature Set: Chat assistance, code explanation, code generation, and context-aware suggestions

Usage Policy

Any AI tools used must connect to Amazon Bedrock and adhere to organizational guidelines regarding AI usage, particularly concerning data privacy and security protocols.

AWS Bedrock Configuration

EU Region Inference Profiles

For compliance with EU data residency requirements, Amazon Bedrock provides inference profile ARNs with the eu. prefix. These profiles ensure that inference requests are processed within EU regions, maintaining full compliance with European data protection standards.

Recommended Model: Claude Sonnet 4.5

• Model ARN: eu.anthropic.claude-sonnet-4-5-20250929-v1:0
• Ensures data stays within EU boundaries
• Full feature parity with global models
• Optimized for code assistance and technical tasks

When configuring your AI code assistant, always use the eu. prefix for model ARNs when working in the EU region:

{
  "model": "eu.anthropic.claude-sonnet-4-5-20250929-v1:0",
  "region": "eu-central-1"
}

Listing Available Bedrock Models in the EU

You can use the following command to list the available models with EU inference profile:

aws bedrock list-inference-profiles --query "inferenceProfileSummaries[?contains(inferenceProfileArn, 'eu-') && starts_with(inferenceProfileId, 'eu.')].inferenceProfileId" --output json | jq -r '.[]'

Tool-Specific Setup

OpenCode

OpenCode is a modern AI-powered code assistant that runs directly in your terminal with native support for Model Context Protocol (MCP) servers.

Installation

Follow the installation instructions on the OpenCode website.

Configuration

OpenCode configuration is managed through the ~/.config/opencode/opencode.json file. Here's a complete configuration example with AWS Bedrock and EU inference profiles:

{
  "$schema": "https://opencode.ai/config.json",
  "provider": {
    "amazon-bedrock": {
      "options": {
        "region": "eu-central-1",
        "profile": "bedrock"
      }
    }
  }
}

For more detailed configuration options, refer to the OpenCode documentation.

Continue.dev

Continue.dev is an open-source IDE plugin that enhances editor functionality with AI capabilities, available for both VSCode and JetBrains IDEs.

Installation

1. Install the plugin of your choice:
   • VSCode
   • JetBrains
2. Open the plugin - you will see a "Get Started" button. You do not need to create an account; you can remain fully local.
3. To configure models, click the gear icon in the top right corner of the Continue panel or directly edit the configuration file located at ~/.continue/config.json.

Adding Chat Models

The configuration file contains a models array where you can specify your preferred language models. Here's a sample configuration for integrating Amazon Bedrock with EU inference profiles:

For JetBrains/VSCode (config.json):

{
  "models": [
    {
      "title": "Claude Sonnet 4.5 (EU)",
      "provider": "bedrock",
      "model": "eu.anthropic.claude-sonnet-4.5-v2:0",
      "region": "eu-central-1"
    }
  ]
}

For VSCode (config.yaml):

models:
  - name: Claude Sonnet 4.5 (EU)
    provider: bedrock
    model: eu.anthropic.claude-sonnet-4.5-v2:0
    env:
      region: eu-central-1
    roles:
      - chat
      - autocomplete

You can customize the title/name field to any name you prefer. This name will be displayed in the model selection dropdown within the Continue interface.

Adding an Autocomplete Model

You can also add an autocomplete model to your configuration for inline code suggestions:

{
  "models": [
    {
      "title": "Claude Sonnet 4.5 (EU)",
      "provider": "bedrock",
      "model": "eu.anthropic.claude-sonnet-4.5-v2:0",
      "region": "eu-central-1"
    }
  ],
  "tabAutocompleteModel": {
    "title": "Claude Sonnet 4.5 Autocomplete",
    "provider": "bedrock",
    "model": "eu.anthropic.claude-sonnet-4.5-v2:0",
    "region": "eu-central-1",
    "completionOptions": {
      "temperature": 0.2,
      "topP": 0.9,
      "topK": 40,
      "maxTokens": 600
    }
  }
}

For more information about Continue.dev configuration, refer to the official documentation.

Claude Code

Claude Code is Anthropic's official code editor with native AI assistance built-in and support for AWS Bedrock.

Installation

Download Claude Code from the official website.

Configuration

Claude Code configuration is managed through settings. Here's how to configure AWS Bedrock with EU inference profiles:

Configuration file location: ~/.claude/settings.json (macOS)

{
  "$schema": "https://json.schemastore.org/claude-code-settings.json",
  "awsAuthRefresh": "boost auth bedrock",
  "env": {
    "CLAUDE_CODE_ENABLE_TELEMETRY": "0",
    "CLAUDE_CODE_USE_BEDROCK": "1",
    "AWS_REGION": "eu-central-1",
    "AWS_PROFILE": "bedrock"
  },
  "model": "eu.anthropic.claude-sonnet-4-5-20250929-v1:0"
}

Other Compatible Tools

Several other tools support Amazon Bedrock integration:

• AWS Toolkit: IDE extensions with built-in Bedrock integration
• Custom integrations: Many tools support Bedrock through AWS SDK integration

Refer to each tool's documentation for specific configuration instructions.

Daily Usage

Authentication

Authentication with Amazon Bedrock is required prior to utilizing any AI code assistant tool. This authentication can be accomplished through the Boost CLI by executing the command:

boost auth bedrock

Upon successful authentication, your credentials will remain valid for 8 hours.

If you receive an error message stating The security token included in the request is invalid, please re-authenticate by executing the command again.

Best Practices

• Use EU inference profile models (eu. prefix) for all work to ensure compliance
• Re-authenticate when your 8-hour token expires
• Leverage MCP servers (see below) for enhanced context and tool integration

Model Context Protocol (MCP) Servers

Model Context Protocol (MCP) enables AI assistants to access external data sources and tools, providing richer context for code assistance. The platform provides several MCP servers that can be integrated with your AI code assistant tool.

Boost Platform Documentation

The Boost Platform Documentation MCP provides AI assistants with access to internal platform documentation, enabling context-aware suggestions based on organizational standards and practices.

You need to connect to the VPN to use this MCP server

{
  "mcp": {
    "boost-docs": {
      "type": "remote",
      "url": "https://boost-ai.eks-eu-central-1-private.prod.platform.carloop.dev/mcp"
    }
  }
}

{
  "mcpServers": [
    {
      "name": "Boost Platform Documentation",
      "type": "streamable-http",
      "url": "https://boost-ai.eks-eu-central-1-private.prod.platform.carloop.dev/mcp"
    }
  ]
}

mcpServers:
  - name: Boost Platform Documentation
    type: streamable-http
    url: https://boost-ai.eks-eu-central-1-private.prod.platform.carloop.dev/mcp

{
  "mcpServers": {
    "boost-platform-documentation": {
      "type": "http",
      "url": "https://boost-ai.eks-eu-central-1-private.prod.platform.carloop.dev/mcp"
    }
  }
}

GitLab MCP

The GitLab MCP server enables AI assistants to interact with GitLab repositories, issues, merge requests, and pipelines directly from your code assistant.

Prerequisites: You need a GitLab Personal Access Token with appropriate permissions. Create one at GitLab Settings.

{
  "mcp": {
    "gitlab": {
      "enabled": true,
      "type": "local",
      "command": ["npx", "-y", "@zereight/mcp-gitlab"],
      "environment": {
        "GITLAB_PERSONAL_ACCESS_TOKEN": "{env:GITLAB_TOKEN}",
        "GITLAB_API_URL": "https://gitlab.carloop.dev/api/v4",
        "USE_GITLAB_WIKI": "false",
        "USE_MILESTONE": "false",
        "USE_PIPELINE": "true"
      }
    }
  }
}

{
  "mcpServers": [
    {
      "name": "GitLab",
      "type": "local",
      "command": "npx",
      "args": ["-y", "@zereight/mcp-gitlab"],
      "env": {
        "GITLAB_PERSONAL_ACCESS_TOKEN": "your-gitlab-token-here",
        "GITLAB_API_URL": "https://gitlab.carloop.dev/api/v4",
        "USE_GITLAB_WIKI": "false",
        "USE_MILESTONE": "false",
        "USE_PIPELINE": "true"
      }
    }
  ]
}

{
  "mcpServers": {
    "gitlab": {
      "command": "npx",
      "args": ["-y", "@zereight/mcp-gitlab"],
      "env": {
        "GITLAB_PERSONAL_ACCESS_TOKEN": "your-gitlab-token-here",
        "GITLAB_API_URL": "https://gitlab.carloop.dev/api/v4",
        "USE_GITLAB_WIKI": "false",
        "USE_MILESTONE": "false",
        "USE_PIPELINE": "true"
      }
    }
  }
}

Tempo Integration & Production MCPs

The Tempo MCP servers provide access to distributed tracing data from the platform's observability infrastructure, enabling AI assistants to help debug and analyze trace data.

Two environments are available:

• tempo-integration: Integration environment traces
• tempo-production: Production environment traces

Credentials: Basic authentication credentials can be retrieved from 1Password (already base64 encoded).

{
  "mcp": {
    "tempo-integration": {
      "type": "remote",
      "url": "https://tempo-gateway.eks-eu-central-1-private.int.shared.carloop.dev/api/mcp",
      "headers": {
        "Authorization": "Basic <base64-encoded-credentials-from-1password>"
      }
    },
    "tempo-production": {
      "type": "remote",
      "url": "https://tempo-gateway.eks-eu-central-1-private.prod.shared.carloop.dev/api/mcp",
      "headers": {
        "Authorization": "Basic <base64-encoded-credentials-from-1password>"
      }
    }
  }
}

{
  "mcpServers": [
    {
      "name": "Tempo Integration",
      "type": "streamable-http",
      "url": "https://tempo-gateway.eks-eu-central-1-private.int.shared.carloop.dev/api/mcp",
      "headers": {
        "Authorization": "Basic <base64-encoded-credentials-from-1password>"
      }
    },
    {
      "name": "Tempo Production",
      "type": "streamable-http",
      "url": "https://tempo-gateway.eks-eu-central-1-private.prod.shared.carloop.dev/api/mcp",
      "headers": {
        "Authorization": "Basic <base64-encoded-credentials-from-1password>"
      }
    }
  ]
}

{
  "mcpServers": {
    "tempo-integration": {
      "type": "http",
      "url": "https://tempo-gateway.eks-eu-central-1-private.int.shared.carloop.dev/api/mcp",
      "headers": {
        "Authorization": "Basic <base64-encoded-credentials-from-1password>"
      }
    },
    "tempo-production": {
      "type": "http",
      "url": "https://tempo-gateway.eks-eu-central-1-private.prod.shared.carloop.dev/api/mcp",
      "headers": {
        "Authorization": "Basic <base64-encoded-credentials-from-1password>"
      }
    }
  }
}

Resources

• Amazon Bedrock Documentation
• OpenCode Documentation
• Continue.dev Documentation
• Model Context Protocol Specification
• ⚡️Lightning Talks⚡️
  • 🎥 Recording
  • 🎞️ Slides