Skip to content

ONE-PAGER.md

Latest commit

 

History

History
253 lines (191 loc) · 8.22 KB

ONE-PAGER.md

File metadata and controls

253 lines (191 loc) · 8.22 KB
Raw

grok-faf-mcp: First MCP Server for Grok

v1.0.0 | FAST⚡️AF - Zero Friction, Instant Access

⚠️ REFERENCE ONLY — superseded launch doc (May 2026). Describes v1.0.0 (17 tools, HTTP-SSE, "forked from faf-mcp"). Does NOT reflect current state: v1.5.3 — refresh_faf, .fafm memory, non-destructive interop, Mk4 WASM scoring, SSE retired, 14 hosted / 59 local tools. Kept for history; do not share as the current one-pager.

What It Is

The first MCP (Model Context Protocol) server built specifically for Grok/xAI, deployed on Cloudflare Workers with URL-based access. No installation required—point your Grok integration at the endpoint and get instant access to 17 MCP tools + 14 bundled commands for universal AI context via IANA-registered .faf format.

Production URLs:

Core Capabilities:

  • 17 MCP tools (score, init, sync, validate, formats, etc.)
  • 14 bundled CLI commands (bi-directional sync, drift detection, framework discovery)
  • HTTP-SSE transport (Server-Sent Events)
  • Global Edge deployment via Cloudflare network
  • IANA-registered format support (application/vnd.faf+yaml)

Why It Exists

The Origin Story: Grok asked for MCP on a URL. This is the answer—delivered with Elon-style execution velocity.

While claude-faf-mcp (6.7k downloads) and faf-mcp (universal MCP) serve the broader ecosystem through local installation, xAI's Grok integration demanded a different approach:

The Requirements:

  1. URL-based access - No npm install, no local runtime, no configuration files
  2. Production infrastructure - Not a demo, not a prototype—battle-tested from day one
  3. Global availability - Cloudflare Workers = 300+ edge locations, sub-50ms latency globally
  4. Zero friction - Add URL to config, start using tools immediately

The xAI Context: When Grok needs to understand project context, it shouldn't require users to install dependencies, configure environments, or debug local servers. It should just work—instantly, everywhere, every time.

grok-faf-mcp is that infrastructure.

Killer Feature

Zero-Installation MCP Integration

Traditional MCP workflow (local install):

// 1. Install package
npm install -g some-mcp-server

// 2. Configure MCP client
{
  "mcpServers": {
    "server": {
      "command": "npx",
      "args": ["-y", "some-mcp-server"]
    }
  }
}

// 3. Debug Node.js runtime issues
// 4. Update when versions change

grok-faf-mcp workflow (URL-based):

Point to: https://mcpaas.live/grok/mcp/v1

Done.

What this eliminates:

  • ❌ No npm install required
  • ❌ No Node.js version conflicts
  • ❌ No local server management
  • ❌ No "works on my machine" debugging
  • ❌ No update cycles (Cloudflare deploys transparently)

What this enables:

  • ✅ Instant access from any MCP client
  • ✅ Works in sandboxed/restricted environments
  • ✅ Always running latest version
  • ✅ Global CDN = sub-50ms latency worldwide
  • ✅ Zero maintenance for end users

Value to xAI / AI Era

For xAI Teams:

  • First-Mover Advantage - grok-faf-mcp is the first MCP server built specifically for Grok (not adapted, not ported—built for)
  • Production-Ready - Forked from faf-mcp v1.1.1, inherits championship testing standards and 19ms average execution speed
  • FAST⚡️AF Branding - Aligns with Elon's execution philosophy: ship fast, ship right, no bureaucracy
  • xAI Strategic Positioning - When Grok asks for MCP, this is the reference implementation

For the AI Era:

  • Infrastructure Precedent - Demonstrates that AI tooling can be URL-native, not just installable packages
  • Ecosystem Velocity - Developers ship Grok integrations in minutes (not hours debugging local setups)
  • Standards Alignment - Built on IANA-registered .faf format + Anthropic MCP spec = interoperable foundation
  • Global Accessibility - Cloudflare Workers means MCP tools work in 300+ edge locations worldwide

Strategic Position:

The Bigger Picture: AI context formats (like .faf) only matter if AI agents can access them. grok-faf-mcp proves that access can be:

  1. Instant - No installation, no configuration
  2. Global - Works everywhere via CDN
  3. Reliable - Production Cloudflare Workers infrastructure
  4. Standard - MCP protocol + IANA format

This combination transforms .faf from "interesting spec" to production infrastructure.

The xAI Narrative: When Grok evaluates project context, it needs three things:

  1. Format - IANA-registered .faf (application/vnd.faf+yaml)
  2. Protocol - MCP (Model Context Protocol)
  3. Infrastructure - grok-faf-mcp (URL-based server)

Together, these create the first complete stack for AI-native context evaluation built specifically for Grok.

Performance & Heritage

Performance Metrics:

Execution Speed:      19ms average
Fastest Command:      1ms (formats)
CLI Speedup:          16.2x faster
MCP Tools:            17 total
Bundled Commands:     14 total
Transport:            HTTP-SSE
Platform:             Cloudflare Workers
Status:               Production

Built on Proven Foundation:

  • Forked from faf-mcp v1.1.1
  • Inherits 17 MCP tools, 14 bundled commands
  • Inherits 16.2x performance gains over previous versions
  • Inherits championship testing standards
  • Inherits IANA-registered format support

Why Separate Package?

  • Grok asked for URL-based MCP (not local install)
  • First-mover positioning for xAI ecosystem
  • Platform-specific branding (FAST⚡️AF)
  • Download count metrics for FAF ecosystem growth
  • Elon-style execution (ship what was asked for, ship it fast)

Quick Start

Option 1: URL-Based (Instant)

Add to MCP client config:
MCP Endpoint: https://mcpaas.live/grok/mcp/v1

Option 2: Local Install (Optional)

npm install -g grok-faf-mcp

# Add to MCP config:
{
  "mcpServers": {
    "grok-faf": {
      "command": "npx",
      "args": ["-y", "grok-faf-mcp"]
    }
  }
}

Then start every prompt with "Use FAF":

Use FAF to initialize my project
Use FAF to score my AI-readiness
Use FAF to sync project context

Technical Architecture

Deployment Stack:

  • Platform: Cloudflare Workers
  • Runtime: Node.js 18+
  • Transport: HTTP Server-Sent Events (SSE)
  • Protocol: MCP via @modelcontextprotocol/sdk
  • Format: IANA-registered application/vnd.faf+yaml

Key Files:

  • mcpaas-cf - Cloudflare Workers MCP server (Streamable HTTP, served at mcpaas.live/grok/mcp/v1)
  • wrangler.toml - Cloudflare Workers config (custom domain grok.faf.one)
  • package.json - grok-faf-mcp package config (inherits faf-cli tools)

Why Cloudflare?

  • Global Edge network (300+ locations)
  • Zero-config deployments
  • Automatic HTTPS
  • Built-in monitoring
  • Transparent updates (users always run latest version)

Community & Ecosystem

Official Links:

FAF Package Family:

  • claude-faf-mcp - 6.7k downloads (Claude Desktop-specific)
  • faf-mcp - 800+ downloads (Universal MCP server)
  • grok-faf-mcp - NEW (Grok/xAI-specific, URL-based)

Related Tools:

  • faf-cli - Command-line interface for FAF operations
  • xai-faf-zig - Native Zig CLI (10,000 scores/second)
  • xai-mcp-server - Rust MCP server (300/300 tests passing)

The Bottom Line

Grok asked for MCP on a URL.

Not a package. Not a binary. A URL.

grok-faf-mcp delivered—first to ship, zero friction, production-ready from day one.

This is championship engineering:

  • Standards that work (IANA + MCP)
  • Code that ships (Cloudflare production)
  • Experiences that resonate (instant access, no setup)

Built for Grok. Built for Speed. Built Right.

Built with: Node.js 18+ | Cloudflare Workers | MCP Protocol | IANA Format Status: Production (https://grok.faf.one/) Philosophy: FAST⚡️AF - First to Ship, Zero Friction Repository: https://github.com/Wolfe-Jam/grok-faf-mcp