[media-types] Real-world experience with application/vnd.faf +yaml (AI context format)
"James@WolfeJames" <[email protected]>
| Newsgroups | gmane.ietf.types |
|---|---|
| Message-ID | <[email protected]> |
Hello, I'm James Wolfe, the registrant of the application/vnd.faf+yaml media type (registered October 30, 2025). I wanted to share some implementation experience that may be relevant to the working group. Background FAF (Foundational AI-Context Format) is a YAML-based format designed to provide persistent project context to AI coding assistants. The format addresses a specific gap: AI tools like Claude, Grok, and GPT lack standardized ways to receive project-specific context that persists across sessions. Implementation Data (as of December 2025) - MCP server merged into Anthropic's official registry (PR #2759, October 17, 2025) - npm ecosystem total: 18,000+ downloads (CLI, MCP server, related tooling) - Chrome extension: Available in Chrome Web Store - Active integrations: Anthropic Claude, xAI Grok, Google Gemini, OpenAI Codex, Cursor, various IDEs Technical Observations The +yaml suffix has worked well for our use case. AI systems parse the format reliably, and the human-readable nature of YAML aligns with our requirement that both humans and AI can read/edit the same file. One area where additional guidance might be useful: media types for AI-specific formats. As AI tooling grows, we may see more formats designed for machine consumption that also need human readability. FAF sits in this intersection. Happy to answer any questions about our implementation experience or provide additional technical details. Regards, James Wolfe https://faf.one https://wolfejam.dev ORCID: https://orcid.org/0009-0007-0801-3841 _______________________________________________ media-types mailing list -- [email protected] To unsubscribe send an email to [email protected]