← All posts

The simplest HTTP MCP server you can build on AWS

November 24, 2025

#generative-ai-tools#serverless#mcp-server#aws-lambda

TL:DR Built a simple AWS application as MCP as a Resume using Lambda invoke URL and Dynamo. Promoting myself with new tech. Allowing 2 ways communication between myself and roles. Using Serverless AWS tools and MCP library.

My MCP Resume: https://bradwebb101.com/about/mcp

Intro:

Are you tired of ignoring recruiters on LinkedIn for a ‘Exciting’ opportunity for a fixed term contract at half your salary, in a technology you have no experience in? I figured I would build a Resume in Model Context Protocol that will allow Recruiters to use AI to assess my suitability for a role before they reach out to me with their unsolicited offers.

When writing this article I found my self day dreaming about the film ‘American Psycho’ with the famous Business Card Scene. Patrick Bateman sliding his business card across the table talking about embossing, the cringe that one must feel. This is that story, just as a ‘Software Developer’ instead of ‘Finance Bro’ and my creation of my ‘MCP as a Resume’.

Before we begin here are the disclaimers:

  • I DO NOT recommend this for a production system.
  • If you want to make this into a production app, you will need to add Cognito Auth as well as a more production ready web access setup (CDN, WAF, Cognito, API Gateway/Load Balancer)
  • If you want to replicate this code, be aware that the lack of Auth means people can spam the endpoint. Could cost you money as this is on the public internet, without any auth. I have cost monitoring on, if the cost of having this endpoint available goes over $5-$10 a month, I will be pulling it down.

A bit about Lambda invoke URL’s:

Lambda Invoke URLs are a lightweight way to expose a Lambda function over HTTPS without configuring API Gateway or additional infrastructure. They’re ideal for prototyping, internal tooling, development environments, and simple event-style integrations where security exposure is minimal and traffic patterns are predictable.

Lambda Invoke URLs have important limitations that prevent them from meeting enterprise-grade production requirements. They do not support Web Application Firewall (WAF) protection, no geoblocking, no fine-grained routing, no usage plans, limited authentication options, and lack advanced threat mitigation, caching, throttling governance, or compliance controls expected in corporate environments.

In short, Lambda Invoke URL’s are good for prototyping, consider migrating to a more robust infrastructure if you want to move the application into production.

So what did I build?

Infrastructure diagram

Libraries and Github Repo

For this lambda function, to simplify the design. I used AWS labs MCPLambdaHandler.

https://github.com/awslabs/mcp/blob/main/src/mcp-lambda-handler/README.md

All the code is available here. Please note Route53 routes + Cloudfront infra was created in a different stack. Either add your own route53 + Cloudfront in the Stack, or use the lambda invoke url as your server address. Find the code in my repo

GitHub - BradWebb101/mcp-as-a-resume: This repo is a Lambda Invoke URL MCP Server. It has public info about my coding and skillset. Maar (MCP as a Resume).

The MCP link again if you missed it https://bradwebb101.com/about/mcp

Accessing like a website

As MCP is built off Post Requests so I added a Get endpoint to the Lambda function, to give instructions as to how to use MCP.

What a GET request returns

Accessing it with the MCP protocol

If you put the URL in MCP inspector you can interact with the server to confirm it is working.

The MCP server has 5 tools

  • get_contact_info — MCP Read. Returns contact information, email, Linkedin medium ect
  • get_articles — MCP Read. Returns recent Medium articles
  • get_skillset — MCP Read. Returns data on my current skillset
  • get_github_repos — MCP Read. Returns 5 most recent Github repos
  • send_email_to_brad — MCP Write. Sends me an email (Its to a live email, so looking forward to the spam)

Interacting with the server

I have added the MCP server to ChatGPT for testing.

Connecting MCP server to ChatGPT

First chat:

First Chat, ‘Who is Brad?’

So far so good, second attempt:

Second chat, what Icode

MCP Write; Still the wild west

I am still skeptical about the use of MCP Write in a LLM world. Read works well but when you have a non-deterministic world writing into a deterministic one (such as a DataBase) opens up a world of problems. For this article I added a Write endpoint to play with, an Email service to me.

Sending me an email

The emails will get to me, so I am curious as to what i will receive.

Conclusion:

This article is a little bit of fun building a MCP resume. If you are a recruiter you can use my tool, I will probably say no to a role as I am very happy in my current position. Be the first recruiter to embrace MCP though.

I built it with the simplest infrastructure possible using MCP tools and AWS Lambda Functions. This is not production ready but a quick and simple way of testing a hypothesis. I think speeding up development to be able to test something in practice is the key for getting value out of AI. It’s not a magic box, it just has has a ‘particular set of skills’.

So with that, I officially declare a new genre for the modern tech professional: MCP as a Resume — MaaR.