# Grok FAF [Health: Active]

**Category:** 🧠 Knowledge & Memory  
**Repository:** https://github.com/Wolfe-Jam/grok-faf-mcp  
**GitHub Stars:** 19  
**npm Downloads (last month):** 2087  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/grok-faf

## Description
Persistent project context for xAI Grok. IANA-registered .faf format.

## Tools
Capabilities this server exposes over MCP:

- **faf_init** — Create project.faf from your project
- **faf_auto** — Auto-detect stack and populate context
- **faf_score** — AI-readiness score (0-100%) with breakdown
- **faf_status** — Check current AI-readability
- **refresh_faf** — Re-ground on the live `.faf` — re-read + re-score, report drift, return fresh DNA (drift → refresh → re-grounded). **Requested by Grok.
- **refresh_fafm** — Re-ground on the live `.fafm` memory layer for one or more souls. Returns a stamped delta (added/updated facts) by default; `verbatim: true` for full content. Read-only · always stamped. Sister to `refresh_faf` for the RAM/memory layer in the vROM/RAM model. **Built for Grok, by request.
- **refresh_blend** — The baked-in two-intensity refresh (Cmd+R / Cmd+Shift+R analog). `mode: "blend"` (default) fires `refresh_faf` (light) + `refresh_fafm` (delta); `mode: "nuke"` fires both at hard intensity. Blend is **BAKED IN, NOT a dial** — both layers always fire; mode only affects fafm intensity.
- **faf_orchestrate_recommendation** — The heavy orchestrator. Reads current substrate state, composes the full 1.5 library substrate (drift detection · CheckID · repeat-offender · take-a-hint · refresh history), returns a structured `Recommendation` with `recommend`, `severity`, `summary`, `reason`, and a rich `hints` object including…
- **faf_get_orchestration_policy** — Pure introspection of the effective policy WITHOUT running the orchestrator. Returns `{ tier, thresholds, source, overrides_applied }` — what aggressiveness tier the next orchestration call would use, and whether it came from defaults or a `.faf:orchestration:` override. No drift detection · no sig…
- **faf_sync** — Sync .faf → CLAUDE.md
- **faf_bi_sync** — Bi-directional .faf ↔ platform context
- **faf_trust** — Validate .faf integrity
- **faf_read** — Read any file
- **faf_write** — Write any file
- **faf_list** — Discover projects with .faf files
- **rag_query** — RAG-powered context retrieval
- **rag_cache_stats** — RAG cache statistics
- **rag_cache_clear** — Clear RAG cache
- **grok_go_fast_af** — Auto-load .faf context for Grok

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `bunx` (confidence: high):

```json
"mcpServers": {
  "grok-faf": {
    "command": "bunx",
    "args": ["faf"]
  }
}
```

## Documentation

## What Grok FAF MCP server does

The Grok FAF MCP server connects xAI Grok to persistent project context stored in `.faf` files. A project file can describe the project name, goals, language, stack, runtime, database, testing setup, and human context in a machine-readable YAML format. The format is identified in the repository as IANA-registered `application/vnd.faf+yaml`.

The Grok FAF MCP server exposes tools for creating, inspecting, scoring, validating, and updating this context. `faf_auto` detects a project stack, creates context, and scores the result. `faf_score` returns an AI-readiness score from 0 to 100 with a breakdown, while `faf_status` checks the current readability state. `faf_trust` validates `.faf` integrity, and `faf_list` discovers projects containing `.faf` files.

Context can be synchronized with `CLAUDE.md` using `faf_sync`, or with platform context through `faf_bi_sync`. The server also provides file access through `faf_read` and `faf_write`, plus RAG retrieval and cache controls through `rag_query`, `rag_cache_stats`, and `rag_cache_clear`. `grok_go_fast_af` loads `.faf` context for Grok.

## How it works

The hosted deployment is configured as an MCP URL. The README provides `https://mcpaas.live/grok/mcp/v1` as the endpoint and shows it in Grok's `~/.grok/config.toml`. After restarting the Grok TUI, or refreshing MCPs with `/mcps r`, the configured tools become available. The hosted service runs on Cloudflare Workers and uses a WebAssembly scoring engine described as the same engine used by the npm package.

For local use, the repository documents a Homebrew installation command. The local package uses stdio, while the hosted option avoids a local subprocess and serves the MCP endpoint from the edge. The README also identifies a Smithery gateway for the project.

Refresh tools support the `.faf` project layer and the `.fafm` memory layer. `refresh_faf` rereads and rescoring the live file, reports drift, and returns refreshed project data. `refresh_fafm` returns a stamped delta by default, with a `verbatim` option for full memory content. `refresh_blend` runs both refresh paths, using either the default blend mode or a harder nuke mode. `faf_orchestrate_recommendation` evaluates substrate state and returns a structured recommendation with severity, reasons, and hints; `faf_get_orchestration_policy` inspects the policy without running that orchestration.

## Setup and configuration

For hosted Grok use, add the following configuration:

```toml
[mcp_servers.grok-faf-mcp]
url = "https://mcpaas.live/grok/mcp/v1"
```

The repository also documents this local installation:

```bash
brew install wolfe-jam/faf/grok-faf-mcp
```

A `/info` endpoint is available at the hosted URL for checking endpoint details, protocol versions, engine information, and the tool list.

## Tools and capabilities

The Grok FAF MCP server includes these documented capabilities:

- Initialize, auto-detect, score, inspect, validate, read, write, list, and synchronize project context.
- Re-ground `.faf` and `.fafm` layers, combine refresh operations, and inspect orchestration policy.
- Produce orchestration recommendations based on current substrate state.
- Query RAG context and inspect or clear its cache.
- Load `.faf` context for Grok with `grok_go_fast_af`.

Some refresh and orchestration operations append receipt files to the project folder.

## Limitations and notes

The hosted endpoint and the local package are described as using an always-33 scoring engine, so scores may differ from other scoring implementations. The hosted deployment exposes a smaller default tool set, with extended utilities available through the documented `FAF_TOOLS=all` setting for local use. Only `rag_query` is identified as reaching the network for xAI Collections.

Write-capable tools can change project files. In particular, the README identifies `faf_init` with `force` and `faf_write` among operations that can overwrite or delete data; review client confirmation behavior before allowing automated writes. Refresh and orchestration calls can also create receipt files even when their primary purpose is reading or analysis.

The Grok FAF MCP server is aimed at projects that can maintain `.faf` context and at MCP clients able to use the documented URL or local stdio configuration.

_Full upstream README: https://allmcps.com/mcp/grok-faf/readme_