-
-
Architecture Diagram
Inspiration
As agentic AI grows in popularity, agent capabilities have become a cornerstone of modern AI development. In today's landscape, MCP (Model Context Protocol) servers provide essential capabilities to AI agents.
However, when developing agent workflows, using actual MCP servers can be problematic since agents might cause unintended side effects during testing and development.
The practical solution is to create mock MCP servers for development purposes. Unfortunately, building these mock servers from scratch is time-consuming and technically challenging.
What it does
Similar to how mock APIs with mock data are essential for full-stack application development, the agentic development ecosystem lacks a standard solution for mocking MCP servers.
This project fills that gap.
It provides a lightweight, configurable Mock MCP Servers on demand
How we built it
MockMCP is a fully AWS-hosted serverless solution built around two core Lambda functions. The first handles user authorization and authentication, while the second manages the complete MCP server lifecycle - from creation to execution.
Users authenticate through AWS Cognito, gaining access to create MCP servers via our Amplify-hosted frontend. All requests - both from users and MCP clients - flow through API Gateway, where our Authorizer Lambda validates and authorizes every interaction before forwarding approved requests to the Server Lambda and generates a policy for the API Gateway.
The entire system operates under the MockMCP.com domain with DNS managed by Route 53. User-uploaded images and MCP server outputs are securely stored in S3 buckets, while the complete state of every MCP server is tracked in a DynamoDB table that the Server Lambda connects to as needed.
This architecture ensures scalability, security, and reliability while maintaining the simplicity that makes MockMCP so effective for rapid development.
Challenges we ran into
I heard about this hackathon at the last stage, so it was challenging to develop the platform in the limited time remaining. However, I was able to complete it successfully. Another difficult part was managing the entire backend infrastructure using SAM, but I also managed to get everything working.
Accomplishments that we're proud of
It’s really great that I was able to add text/JSON, image, and custom JS output workflows as well. Initially, I didn’t expect to have enough time to implement the custom JS workflow, but now that’s done too. Developers can define conditional outputs using JavaScript and also verify their code.
What we learned
I learned many things about the internal structure of the MCP protocol. Even though I had used it a lot before, I never really wanted to dive into its internal structure earlier. The best part is that Cursor was helpful during development, but it also caused a lot of trouble. What I learned from this is that I can never fully trust Cursor for making architectural decisions. However, once I make the decisions myself, Cursor can implement them properly.
What's next for MockMCP
The actual subscription based pricing capabilities are not yet developed, so the current platform allows developers use it forfree, but as the next step i will implement the pricing structure and subscription based payment handling
Log in or sign up for Devpost to join the conversation.