claude-faf-mcp and claude-code-mcp
These two tools appear to be **ecosystem siblings**, specifically different implementations or components within the same Claude MCP (Master Control Program) ecosystem, with "KunihiroS/claude-code-mcp" focusing on connecting with Claude code via local commands and "Wolfe-Jam/claude-faf-mcp" providing an Anthropic MCP server for the IANA-registered `.faf` format.
About claude-faf-mcp
Wolfe-Jam/claude-faf-mcp
Anthropic MCP server for .faf — 33 tools, IANA-registered format. npm: claude-faf-mcp. MCP Registry #2759
Provides 33 MCP tools for creating and managing `.faf` files—a YAML format that captures project context (who it's for, what it does, why it exists) to optimize AI understanding across sessions. Features bi-directional and tri-directional sync with `CLAUDE.md` and `MEMORY.md`, auto-detection of project stacks, AI-readiness scoring (0-100%), and interoperability with other LLM platforms (Gemini, Grok, Cursor) through a shared `.faf` specification. Runs entirely locally with zero dependencies, averaging 19ms per tool execution, and uses stdio transport for Claude Desktop integration.
About claude-code-mcp
KunihiroS/claude-code-mcp
MCP Server connects with claude code local command.
Exposes Claude Code CLI capabilities (explain, review, fix, edit, test code) as MCP tools through stdio-based JSON communication, using Base64 encoding to safely handle special characters in natural language inputs. Implements request routing via `child_process.spawn` to execute templated or custom Claude CLI commands and return results to MCP hosts. Integrates with Claude Desktop and other MCP hosts, supporting Node.js 18+ with configurable logging and environment-based CLI path configuration.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work