Software

AWS Builds MCP Bridge for AgentCore AI Agent

AWS developed an MCP bridge to connect its AgentCore-hosted AI agent with local MCP tools, enabling secure access to user files like Excel spreadsheets.

"Midjourney: 25 variations of an industrial warehouse" by kevin dooley is licensed under CC BY 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/2.0/.

AWS announced the development of an MCP bridge to connect its AgentCore-hosted AI agent with local MCP tools. This bridge allows the AI agent to access user files, such as Excel spreadsheets, stored on local machines. The Model Context Protocol (MCP) is an open-source standard introduced by Anthropic in November 2024 to standardize how AI models connect to external data and tools. The bridge enables secure communication between the remote MCP client and the local MCP server, supporting use cases such as financial analysis where users work with local files. The solution is fully self-hosted on AWS, allowing users to deploy their own models and custom tool servers. The bridge uses WebSocket and native messaging to tunnel MCP messages, ensuring secure and efficient data transfer. The complete source code is available on GitHub, and the MCP Bridge Demo extension provides a practical example of this integration. The architecture includes an AgentCore runtime, a browser extension, an MCP Bridge, and a local MCP server, working together to relay messages between the cloud and local environments. Source: awsml

AWS built a production-grade AI assistant for finance that has seen over 41,000 conversations within a year since launch. The agent is deployed on Amazon Bedrock AgentCore and uses MCP servers that run on a user’s local machine. The bridge connects the remote MCP client to the local MCP server by tunneling MCP messages over WebSocket and native messaging. This allows the AI agent to access local files while maintaining security and performance. The solution supports financial managers and analysts who primarily work with Excel and local files, enabling them to use centrally deployed AI agents to act on those files while drawing context from their browser. Source: awsml

The MCP Bridge works by translating between the native messaging envelope format and raw MCP JSON-RPC. The browser extension acts as a bidirectional relay, forwarding MCP messages between the AgentCore runtime and the MCP Bridge. The MCP Bridge is a FastMCP proxy running on the user’s local machine and is spawned by the browser through the native messaging host registration. The bridge ensures secure communication by using the stdio transport mechanism, which is suitable for local server-client interactions. The complete architecture includes an AgentCore runtime, a browser extension, an MCP Bridge, and a local MCP server, all working together to relay messages between the cloud and local environments. Source: awsml

Source: awsml

Key points

AWS developed an MCP bridge to connect its AgentCore-hosted AI agent with local MCP tools.
The bridge enables secure access to user files like Excel spreadsheets stored on local machines.
The Model Context Protocol (MCP) is an open-source standard introduced by Anthropic in November 2024.
AWS built a production-grade AI assistant for finance that has seen over 41,000 conversations within a year since launch.
The MCP Bridge translates between the native messaging envelope format and raw MCP JSON-RPC.
The browser extension acts as a bidirectional relay, forwarding MCP messages between the AgentCore runtime and the MCP Bridge.
The MCP Bridge is a FastMCP proxy running on the user’s local machine and is spawned by the browser through the native messaging host registration.

Source: AWS Machine Learning Read the original →

WRITTEN BY

Theo Almeida

AI Software & Developer Tools

Theo covers AI software, developer tools, frameworks, and the platforms builders use every day.