DEV Community

wolfejam.dev
wolfejam.dev Subscriber

Posted on

From Desktop Frustration to Internet Standard in 3 Months

#webdev #ai #productivity #opensource

.faf simplifies AI context - from chaos to clarity

.faf Officially Registered by IANA as application/vnd.faf+yaml

From Desktop Frustration to Internet Standard in 3 Months

IANA (Internet Assigned Numbers Authority) officially registered .faf as application/vnd.faf+yaml - the first AI context format with official MIME type designation. It sits alongside PDF, JPEG, XML, JSON, and HTTP itself as an internet standard.

What is .faf?

.faf stands for Foundational AI-context Format -

A file format for persistent AI project context. One project.faf file (sits with package.json and README) in your repo gives AI assistants complete project understanding - tech stack, architecture, dependencies, conventions.

project.faf file in repository structure

Works with Claude Desktop (MCP), Claude Code, OpenAI Codex, Cursor, Gemini CLI, Windsurf, Warp - any AI tool that can read files.

The Problem

Most projects have 40-60% AI-context at best. At 55% you're flipping a coin with AI at every decision. AI assistants don't understand your project, explanations don't persist, context is lost. Projects suffer, run late, many fail. AI can't help you effectively.

The Solution

AI needed a format, it got one. .faf brings your project to a minimum of 85-99% (recommended) AI-context. Simply put: the closer you get to 100%, the more AI can actually help you. At 99/100, AI is optimized.

It's the only format-driven solution for persistent context provided at the foundational layer. It's not claude.md, not an afterthought. They do an important but different job—.faf defines, Claude interprets.

.faf lives in git with your code, survives across sessions and tools. It provides foundational facts to any AI like package.json does for dependencies.

Quadruple Validation

  • IANA Registration (Oct 31, 2025) - Official internet media type
  • Anthropic MCP Registry (Oct 17, 2025) - PR #2759 merged, first persistent project context server
  • Google Chrome Web Store (September 2025) - Extension approved
  • 13k+ downloads - 6.5k CLI + 6.7k MCP (800+/week sustained)

The Real Journey (30 Years)

  • 1995-2025: Building format expertise - Commodore Amiga's .iff, Carpet Industry 3D simulations
  • 2024: Started AI research and development
  • 2025: Started AI format development, immediately saw the context problem and lack of a format
  • August 8, 2025: Format created, first ever .faf
  • September 1, 2025: Developer tool with .faf engine-Mk1
  • September 11, 2025: Google approved Chrome Extension
  • September 16, 2025: MCP server launched (v2.0.0)
  • September 24, 2025: CLI launched (v2.1.0)
  • October 17, 2025: Merged into Anthropic MCP Registry (PR #2759)
  • October 31, 2025: IANA registration

How to Use

Install via npm or Homebrew:

# npm
npm install -g faf-cli

# Homebrew
brew install faf-cli

# Quick start
cd your-project
faf init        # Creates .faf file
faf score       # Check AI-context (0-100%)
Enter fullscreen mode Exit fullscreen mode

For Claude Desktop (MCP), add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "faf": {
      "command": "npx",
      "args": ["-y", "claude-faf-mcp"]
    }
  }
}
Enter fullscreen mode Exit fullscreen mode

30-Year Format Expertise

This wasn't luck - it was three decades of format design experience applied to AI's context problem. From .iff to Carpet Industry simulations to AI context, every format taught me something. When AI needed a format, I got it right immediately because I'd been preparing for 30 years.

30 years collating. 3 months creating! 💥

Resources

Documentation:

CLI Tools:

MCP Servers:

Open source, MIT licensed, free forever.

Persistent, format-driven context is essential. The closer you get to 100% AI-context, the more AI understands your project.

🏎️✨wolfejam.dev

"It's so logical if it didn't exist, AI would have built it itself" — Claude

Top comments (5)

Collapse
 
designzbyoj profile image
DesignzByOJ •

30 years?! Boy that's incredible to have persisted so long, you've come out with an awesome website showcasing it too! Huge respect 👏

Collapse
 
wolfejam profile image
wolfejam.dev •

Thank you so much—this is incredible. My skills are now dedicated to solving the most critical AI problem: zero context.

The AI has no memory and constantly needs reminding via inconsistent documentation.

The solution is a reliable, fact-only Format that is always accessible. That's project.faf, the foundational 'README for AI.'"

🏎️✨wolfe

Collapse
 
carl231 profile image
carl •

honestly, the timing makes sense — everyone’s been fighting the same problem: AI tools just don’t remember project context, and every session feels like starting from zero.

Collapse
 
wolfejam profile image
wolfejam.dev •

before.faf-after.faf

this diagram highlights the issue, this example is Claude but they are all very similar, they have no 'goto' for the facts.

project.faf is like package.json, but for AI context, not dependencies.

Try it out and let me out if you have any questions

I appreciate your comment

Collapse
 
designzbyoj profile image
DesignzByOJ •

True that's been a great feature. Recently I tried sesame ai, even without an account it remembers the conversation after weeks later XD sometimes I'd rather it didn't