-
-
Baseline Check: AI agent (Claude) validates 100% compliant code. Instant confirmation of production readiness via the MCP server
-
AI-Assisted Fix: Agent corrects the code based on our tool's report, boosting the score from 67% to 75% Baseline compatibility
-
MCP Validation: Testing our server directly with the MCP Inspector client to confirm accurate reporting and protocol adherence
Inspiration
Currently, I'm deeply interested in developing AI Agents for coding assistance. Having recently written an article about the Model Context Protocol (MCP), I found it would be particularly relevant to create a utility MCP server.
The goal was to link the power of AI Agents to a critical standard for the modern web: Baseline compatibility. My project positions itself as the bridge between coding AI and web feature reliability.
What it does
The Baseline Checker MCP Server exposes two main tools for MCP clients:
check_baseline (Main tool)
This tool analyzes HTML, CSS, and JavaScript code and provides a comprehensive diagnosis:
- Complete code analysis and compatibility percentage calculation.
- Detailed report that groups features by status: Baseline, Non-Baseline, and Discouraged.
- Clear warnings about potential browser support.
- Recommendations to replace obsolete elements or add fallback solutions.
check_baseline_from_selection
This tool is designed to allow IDEs to automatically retrieve user-selected code
How we built it
This project was built using the following technologies:
- Node.js / TypeScript for MCP server development.
- The
web-featurespackage to access official compatibility and Baseline status data. - The Model Context Protocol (MCP) to establish standard communication with code agents and clients (like MCP Inspector).
Challenges we ran into
The main technical challenge was reconciling the complexity of the MCP protocol with integrating the web-features package to extract relevant information and format it correctly for the agent. I had to use AI to help me unblock and optimize certain parts of the server code.
Accomplishments that we're proud of
I'm very proud to have completed my first Devpost hackathon and successfully delivered a complex project. Developing an MCP server and making it functional using a language I don't yet master perfectly represents a great personal achievement.
What we learned
- I discovered and learned the inner workings of the
web-featurespackage. - I became even more familiar with the Model Context Protocol (MCP) by implementing it concretely, beyond theory.
What's next for Baseline Checker MCP Server
I plan to add other analysis tools to make this MCP server even more interesting. The top priority is to implement Bulk File Analysis.
Bulk File Analysis: A new tool to check multiple files at once (e.g., an entire project folder or several selected files). This is crucial for real-world projects, as developers need a single, overall Baseline score for their entire application.
Built With
- cursor
- kiro
- mcpjam-inspector
- modelcontextprotocol/sdk
- node.js
- typescript
- web-features
Log in or sign up for Devpost to join the conversation.