HUMANS w/AI at Work

FAF Foundation

Stewards of the Foundational AI-context Format

IANA-registered · Universal AI-Context for any AI, human, or team

Get Involved faf.one View Specification
IANA Registered MCP #2759 Chrome Approved 20K+ Downloads

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

"Should become the standard."

— Claude Code, Sep 18 (42 days later → IANA Registered Oct 30, 2025)

Latest News

FIRST Jan 2026

grok-faf-mcp: Three Industry Firsts

First MCP for xAI Grok. First web-based MCP. Fastest growing FAF package with 100+ downloads overnight.

RESEARCH Jan 14, 2026

Academic Whitepaper on Zenodo

Peer-citable technical paper with DOI. Repository used by NASA, NIH, and major universities.

DOWNLOADS

20,000+ npm Downloads

Combined downloads across faf-cli, claude-faf-mcp, faf-mcp, and grok-faf-mcp.

Why FAF Exists

The context problem that every developer faces

Modern development chaos - React, Node, AWS, Claude, Git, Dependencies, Errors, Stress - unified by .faf Project DNA for ANY AI

AI Tools Don't Share Context

Claude doesn't know what you told Gemini. Cursor doesn't know your ChatGPT history. Every tool starts from zero.

You Re-Explain Everything

Stack, conventions, architecture, rules. Every session. Every tool. The same context, typed again and again.

Context Drift is Silent

CLAUDE.md says one thing. Your team's mental model says another. No single source of truth.

The Solution

One format. Every AI. Authored once, works everywhere.

Before

29%

context understood

→

After .faf

99%

context understood

Setup time: 20+ minutes → under 3 seconds

"package.json wasn't built for this, .faf was."

"package.json gives me a list of dependencies, .faf shows me how to use them."

— .faf Inventor + Claude Code

How It Works

.faf sits at the center. Bidirectional conversion to any format.

CLAUDE.md
↔
.faf
↔
GEMINI.md
↕
conductor/

Your project DNA travels with you regardless of which AI you use.

project.faf file sitting alongside package.json and README.md in a project directory

project.faf sits alongside package.json and README.md

"Every project should have one."

— OpenAI Codex CLI (9/10)

Compatible With

Claude Gemini Grok Codex GitHub Cursor Chrome

Before .faf

Skills
Ephemeral
Projects
Session-scoped
MCP
Connection-only
Subagents
Isolated
No Foundation
Components floating independently. Context reinvented per session.

After .faf

Skills
Loads .faf
Projects
Serializes .faf
MCP
Returns .faf
Subagents
Shares .faf
.faf Format Foundation
IANA-Registered Persistent Universal

The Format

YAML-based project DNA that any AI can read

project.faf 
faf: 3.0
name: my-project
description: AI-ready project DNA

stack:
  language: typescript
  framework: next.js
  runtime: node 20

conventions:
  - strict TypeScript
  - server components first
  - Tailwind for styling

ai_context:
  priority: performance
  avoid: any types

The 50/50 Philosophy

Half technical (YAML for AI), half human-focused (markdown for you). Balance by design.

01 Human-Readable

YAML format that developers can read and write. No special tooling required.

02 AI-Optimized

Structured data that AI systems parse instantly. No guessing, no hallucination.

03 Version-Controlled

Lives in your repo. Changes tracked with git. Team context always in sync.

Full Specification

"README evolution for AI era."

— Google Gemini CLI (9.5/10)

Who Uses FAF

Official implementations and community adoption

Official Implementations

MCP
claude-faf-mcp

Anthropic MCP #2759

CLI
faf-cli

8,800+ downloads

EXT
Chrome Extension

Google approved

Projects Using FAF

theblockchain.ai

Production deployment

GALLERY-SVELTE

Svelte component system

FAF-Voice

Voice AI context layer

Validation

11,240+

Tests run

9.3/10

Average rating

<50ms

Processing time

Gold-Code*

100% AI-optimized, persistent project context. Full context. Zero drift. Best code possible.

Using FAF? Let us know to be featured here.

Our Mission

To steward FAF as an open, vendor-neutral standard that enables any AI to understand any project.

Open Standard

FAF is and will always be an open standard. The specification is MIT-licensed, allowing anyone to implement, extend, and build upon it without restriction.

Vendor Neutral

Works with Claude, GPT, Grok, Gemini, and any future AI. No lock-in, no proprietary extensions that fragment the ecosystem.

Community Driven

Evolution of the format is guided by real-world usage from developers, AI companies, and tool makers who use FAF daily.

Add FAF to Your Tool

Three paths to FAF integration

1

Read .faf Files

Parse project.faf to understand project context. Standard YAML - use any parser.

import yaml

with open('project.faf') as f:
    context = yaml.safe_load(f)

# Your AI now knows:
# - Project stack
# - Coding conventions
# - AI-specific instructions
2

Generate .faf Files

Analyze a codebase and output project.faf. Auto-detect stack, frameworks, patterns.

# Use faf-cli for reference
npx faf-cli init

# Or build your own detector
# following the FAF spec schema
3

Export to Formats

Convert .faf to platform-specific formats: CLAUDE.md, GEMINI.md, conductor/, etc.

# Export examples
faf claude export  # → CLAUDE.md
faf gemini export  # → GEMINI.md
faf cursor export  # → .cursorrules

Building an AI tool? We want to help.

Contact for Integration Support

Who It's For

One format, many use cases

For Developers

Write your project DNA once. Export to CLAUDE.md, GEMINI.md, or any format. Zero drift across sessions.

For AI Tools

Add FAF support to your product. IANA-registered format with official SDKs in TypeScript, Python, and Rust.

For Teams

Version-controlled project context. Onboard new devs instantly. Consistent AI interactions across your org.

Resources

Official SDKs, specifications, and tools

📄 Format Specification

The official FAF specification document with examples and validation rules.

Read Specification →

📚 Academic Whitepaper

Peer-citable research paper on CERN/Zenodo. DOI: 10.5281/zenodo.18251362

Read on Zenodo →

📰 Press Release

Overview of FAF format, the problem it solves, and adoption status.

Read Press Release →

🦀 Rust SDK

High-performance Rust implementation. Owns "ai-context" keyword on crates.io.

View on crates.io →

🐍 Python SDK

Full-featured Python implementation with validation. Published on PyPI.

View on PyPI →

📦 TypeScript SDK

Official TypeScript/JavaScript implementation via FAF CLI.

View on npm →

🔧 Claude MCP Server

Official MCP server for Claude Desktop integration. Anthropic MCP #2759.

View on npm →

🚀 Grok MCP Server

First MCP server for xAI Grok. Published on npm.

View on npm →

💻 CLI Tool

Command-line interface for creating and validating FAF files.

Visit fafcli.dev →

Governance

How the Foundation operates

Open Governance

Transparent, community-driven, and focused on long-term sustainability. Decisions are made in the open, on GitHub, with input from the community.

Key governance areas:

  • Specification evolution and versioning
  • Official SDK maintenance
  • IANA registration stewardship
  • Community standards and best practices
  • Enterprise licensing guidance

Licensing

The FAF specification and reference implementations are MIT-licensed, providing maximum flexibility for adoption.

Enterprise options available for organizations requiring:

  • Priority support and SLAs
  • Custom integrations
  • Training and consultation
  • Roadmap influence

How Changes Happen

1

Suggest

Open issue or discussion

→
2

Discuss

Community feedback

→
3

Implement

PR and review

→
4

Release

Version bump

All major decisions go through github.com/Wolfe-Jam/faf

Roadmap

Where FAF is heading

2026 Priorities

  • Native AI platform integrations
  • Extended format schema (v4.0)
  • Enterprise governance features
  • Multi-language SDK parity

How to Influence

Contribute

  • SDKs: TypeScript, Python, Rust
  • Validators and tooling
  • Documentation improvements
  • Platform integrations
View on GitHub

Milestones

Key moments in FAF history

Aug 08 2025

First .faf Created

The very first .faf file was created in fafdev.tools.

Sep 2025

Format Created

FAF specification v1.0 published. Initial TypeScript implementation.

Oct 17 2025

Anthropic MCP

FAF MCP server approved and merged into Anthropic's official MCP ecosystem.

Oct 2025

IANA Registration

application/vnd.faf+yaml registered with IANA. Official MIME type established.

Nov 2025

First MCP for Grok/xAI LIVE

grok-faf-mcp published — first Model Context Protocol server for xAI Grok.

Jan 14 2026

20,000 Downloads NEW

FAF ecosystem surpasses 20,000 npm downloads across faf-cli, claude-faf-mcp, faf-mcp, and grok-faf-mcp.

Jan 14 2026

Academic Whitepaper Published DOI

Peer-citable research paper published on Zenodo. Repository used by NASA, NIH, and major universities.

Jan 2026

Foundation In Formation NOW

Soft launch. Founding sponsors welcome. Building the governance structure.

Founding Sponsors

Building the foundation together

The FAF Foundation is in formation.

We're seeking founding sponsors who believe in open AI context standards. Early supporters shape the roadmap and get prominent recognition as the foundation grows.

Become a Founding Sponsor Join the Discussion

20K+

Downloads

IANA

Registered

8+

Platforms

Join the Foundation

Help shape the future of AI context standards

GitHub Contact