Proposal: Official TC39 MCP Server for Proposals, FAQs, and Contribution Guidelines

[hemanth]

Summary

I propose that TC39 provide an official Model Context Protocol (MCP) server to make TC39 proposals, FAQs, and contribution guidelines more accessible through AI tools.

Background

Model Context Protocol (MCP) is an open standard that enables AI assistants to securely access data sources. An MCP server would allow developers to query TC39 proposals directly from their AI tools (Claude Desktop, VSCode with AI extensions, etc.) rather than manually browsing documentation.

Proposed Features

An official TC39 MCP server could provide:

1. Proposal Management

   • Search proposals by name, description, champion, or author
   • Filter by stage (-1, 0, 1, 2, 2.7, 3, 4)
   • Get detailed proposal information including specification status
   • Compare proposals side-by-side
   • List proposals by champion

2. FAQ Access

   • Query TC39 frequently asked questions
   • Search FAQ topics
   • Get process-related information

3. Contribution Guides

   • Access "How to Contribute" documentation
   • Understand the proposal process
   • Get meeting participation guidelines

Benefits

• Developer Experience: Developers can query "What TC39 proposals are at Stage 3?" directly in their AI tools
• Discoverability: Makes TC39 work more discoverable to the broader JavaScript community
• Real-time Access: Always up-to-date information sourced from the official dataset
• Standardized Interface: Consistent API for all MCP-compatible tools

Proof of Concept

I've built a working implementation as a proof of concept: https://tc39-mcp.deno.dev

The server currently provides:

• Tools for searching, filtering, and comparing proposals
• Resources for accessing individual proposals and summaries
• Prompts for common queries

GitHub: https://github.com/hemanth/tc39-mcp

Implementation Considerations

• Could be deployed on Deno Deploy, Cloudflare Workers, or similar edge platform (free tier sufficient)
• Would use the existing https://tc39.es/dataset/proposals.json as the data source
• Low maintenance - mostly automated updates from the dataset
• Could integrate with tc39.es authentication if needed for write operations

Related Work

• Model Context Protocol specification: https://modelcontextprotocol.io/
• Anthropic's MCP announcement: https://www.anthropic.com/news/model-context-protocol

Would the TC39 team be interested in maintaining an official MCP server, or alternatively, endorsing/linking to community implementations?

See it in action 👇

tc39-mcp-server.mov

[ljharb]

I very much doubt it would be a good idea to have any official solution in this domain, but I look forward to trying out your POC.

[hemanth]

Thanks Jordan! Totally understand the hesitation around "official" in this domain.

A few thoughts on why it's useful:

• Zero maintenance - Just consumes the existing proposals.json. No new data to maintain.
• Read-only - No write ops, minimal risk. A structured API over existing public data.
• Developer discovery - AI tools are increasingly how developers consume documentation. Having accurate TC39 data accessible beats fragmented/hallucinated responses.

Would TC39 be open to just linking to community implementations from tc39.es? That way developers can discover it without any maintenance burden.

Looking forward to your feedback on the mcp server liked above 🙏

[hemanth]

@ljharb any luck?