Harness Score L1 (Documented): measures AI-assisted development harness maturity with harness-score

mcp-me

Your AI assistants don't know who you are. Every time you start a conversation with Claude, Copilot, Cursor, or Windsurf, it's a blank slate — no context about your skills, your projects, your career, or what you care about.

mcp-me fixes that. It creates a structured personal profile that any AI assistant can read via the Model Context Protocol (MCP). Think of it as a digital identity layer for AI — your bio, career, skills, interests, projects, and more, always available to every AI tool you use.

You: "Write me a cover letter for this job"
AI:  (reads your me://career, me://skills, me://projects)
     "Based on your 5 years at Acme Corp, your TypeScript expertise,
      and your open-source work on mcp-me..."

Why mcp-me?

• AI that knows you — Your assistants remember your skills, career, projects, and personality across every conversation
• Auto-generated — Pull data from 329 registered generators (implemented across 44 generator source files) with one command
• Privacy-first — All data stays local in YAML files on your machine. Nothing is sent to any cloud.
• Real-time plugins — 13 live integrations (Spotify now playing, GitHub repos, Last.fm scrobbles) that AI queries on demand
• Extensible — Community-driven generators and plugins. Add a new data source in ~10 lines of code.
• Works everywhere — Claude Desktop, Cursor, Windsurf, Copilot, and any MCP-compatible AI assistant

Installation

Prerequisite: Node.js 20 or later. Verify with node -v.

One-Click Install

Add mcp-me to your AI assistant in one click — no path configuration needed (profile defaults to ~/.mcp-me):

Claude Desktop: Download the latest .mcpb release and double-click to install, or drag it into Claude Desktop → Settings → Extensions.

After installing, initialize your profile:

mcp-me init
mcp-me generate --github your-username

npm (recommended)

npm install -g mcp-me

This makes the mcp-me command available everywhere on your system. No need to clone any repository — npm downloads the package for you.

mcp-me --help
mcp-me init
mcp-me generate --github your-username
mcp-me serve

Alternative — run without installing (via npx):

npx mcp-me --help

npx downloads the package temporarily and runs it. Useful for trying mcp-me once, but slower on repeated use since it re-downloads each time.

Profile location: Commands default to ~/.mcp-me. Override with MCP_ME_PROFILE_DIR or pass an explicit path: mcp-me serve ~/my-profile.

Quick Start

# 1. Initialize your profile (creates YAML templates + .mcp-me.yaml in ~/.mcp-me)
mcp-me init

# 2. Edit the config file — uncomment your sources
code ~/.mcp-me/.mcp-me.yaml

Your .mcp-me.yaml config file:

generators:
  github: your-username
  devto: your-username
  bluesky: handle.bsky.social
  zodiac: aquarius

plugins:
  github:
    enabled: true
    username: your-username

# 3. Generate! Reads sources from .mcp-me.yaml automatically
mcp-me generate

# 4. Start the MCP server
mcp-me serve

CLI flags also work: mcp-me generate --github octocat --devto myuser

All commands work with npx (zero install) or with mcp-me directly if installed globally. The generate command pulls your data from public APIs and auto-populates profile YAML files — no API keys needed for most sources.

Profile directory structure

~/.mcp-me/             → Default profile location
  .mcp-me.yaml        → Configuration (generators + plugins)
  identity.yaml        → Your data (name, bio, contact)
  skills.yaml          → Your data (languages, tools)
  projects.yaml        → Your data (portfolio)
  career.yaml          → Your data (experience)
  interests.yaml       → Your data (hobbies, topics)
  personality.yaml     → Your data (traits, values)
  goals.yaml           → Your data (short/long-term)
  faq.yaml             → Your data (Q&A pairs)

Configure Your AI Assistant

One-Click (Cursor & VS Code)

Use the one-click install badges above. After installing, run mcp-me init to create your profile.

Cursor Plugin (Open Plugins)

mcp-me ships as a full Open Plugins plugin for Cursor. Install from cursor.directory or clone the repo — the plugin auto-detects all components:

Validate the plugin structure: npm run validate:open-plugin

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "me": {
      "command": "npx",
      "args": ["-y", "mcp-me", "serve"]
    }
  }
}

Cursor (manual)

Add to .cursor/mcp.json in your project root:

{
  "mcpServers": {
    "me": {
      "command": "npx",
      "args": ["-y", "mcp-me", "serve"]
    }
  }
}

VS Code (GitHub Copilot)

Add to .vscode/mcp.json in your project root (or in your User Settings for global access):

{
  "servers": {
    "me": {
      "command": "npx",
      "args": ["-y", "mcp-me", "serve"]
    }
  }
}

Tip: To enable it globally (all workspaces), open VS Code Settings (Ctrl+Shift+P → "Preferences: Open User Settings (JSON)") and add the mcp key there instead.

Claude Desktop

Option A — Desktop Extension (.mcpb, recommended):

1. Download mcp-me.mcpb from GitHub Releases
2. Double-click the file, or drag it into Claude Desktop → Settings → Extensions
3. Set your profile directory when prompted (default: ~/.mcp-me)
4. Run mcp-me init if you haven't created a profile yet

Option B — Manual config:

Add to your Claude Desktop config:

{
  "mcpServers": {
    "me": {
      "command": "npx",
      "args": ["-y", "mcp-me", "serve"]
    }
  }
}

Agent Instruction Files

AI agents read special Markdown files in your project to understand how they should behave. Pair them with mcp-me so that every AI session automatically consults your profile — no more repeating your stack, career, or preferences in every conversation.

Quick setup:

# AGENTS.md — for Windsurf, Cline, Codex, etc.
cp "$(npm root -g)/mcp-me/templates/AGENTS.md" ./AGENTS.md

# GitHub Copilot
mkdir -p .github
cp "$(npm root -g)/mcp-me/templates/.github/copilot-instructions.md" .github/copilot-instructions.md

Both templates are included with mcp-me and tell the AI to call ask_about_me whenever context about you would improve the response.

See the Agent Instructions Guide for examples, global setup, and tips on combining with project-specific rules.

Profile Schema

Your profile is a collection of YAML files:

See Schema Reference for full documentation.

MCP Interface

Resources

Static profile data exposed as MCP resources:

• me://identity — Personal identity and contact
• me://career — Professional history
• me://skills — Skills and proficiencies
• me://interests — Hobbies and preferences
• me://personality — Personality traits and values
• me://goals — Personal and professional goals
• me://projects — Portfolio and projects
• me://faq — Frequently asked questions

Tools

• ask_about_me — Free-form question about the user
• search_profile — Keyword search across all profile data

Prompts

• introduce_me — Generate a 2-paragraph introduction
• summarize_career — Summarize career trajectory
• technical_profile — Describe technical skills and stack
• collaboration_fit — Evaluate fit for a project

Generators (selected examples)

Generators run during mcp-me generate to auto-populate your profile from public APIs or user-provided exports. No API keys needed for most sources.

Current counts:

• Registered generators: 329
• Generator source files: 44 single-generator files + 15 batch files (src/generators/)

The table below is a curated list of commonly used generators. The complete source of truth is src/generators/index.ts.

Blogger XML Backup

--blogger-backup imports written content from a Blogger XML export file on disk. It is designed for archived or multi-author blogs where you want to recover your posts even if the site is old, partially offline, or not easily queryable through a public API.

By default it imports all post entries from the file. If the backup contains multiple authors, append ::author1,author2,... after the file path to keep only posts whose author name or email matches one of those values.

Examples:

# Import all posts from a Blogger export
mcp-me generate --blogger-backup ~/Downloads/blog-2026-03-24.xml

# Import only posts written by specific authors/emails
mcp-me generate --blogger-backup "~/Downloads/blog.xml::fernandopalad@gmail.com,fnpaladini@gmail.com,Fernando Paladini"

How to get the export file:

1. Open Blogger and choose your blog.
2. Go to Settings.
3. In Manage blog, click Back up content.
4. Download the XML file.
5. Pass that file path to --blogger-backup.

The generator imports matching posts into projects.yaml with category: article, preserves post labels as tags, and adds summary FAQ entries about the archive. | Activity | --letterboxd <user> | Letterboxd RSS | Films watched, ratings | | Activity | --goodreads <user> | Goodreads RSS + author page | Read books, shelves, reviews, published books | | Activity | --chess <user> | Chess.com API | Rating, stats | | Activity | --lichess <user> | Lichess API | Rating, games | | Activity | --codewars <user> | Codewars API | Rank, honor, languages | | Activity | --leetcode <user> | LeetCode GraphQL | Problems solved, contests | | Activity | --lastfm <user> | Last.fm API | Listening history, top artists | | Activity | --steam <id> | Steam API | Games, playtime | | Activity | --twitch <user> | Twitch API | Stream info | | Activity | --dribbble <user> | Dribbble | Design shots, portfolio | | Activity | --unsplash <user> | Unsplash API | Photos, downloads, collections | | Activity | --exercism <user> | Exercism API | Language tracks, exercises | | Activity | --hackerrank <user> | HackerRank API | Badges, challenges solved | | Activity | --anilist <user> | AniList GraphQL | Anime/manga stats, genres | | Identity | --gravatar <email> | Gravatar API | Bio, linked accounts, photo | | Identity | --keybase <user> | Keybase API | Verified identity proofs |

Want to add a new data source? See the Generator Creation Guide.

Plugins (13 live integrations)

Plugins run during mcp-me serve and provide real-time data to AI assistants on every query.

Enable plugins in .mcp-me.yaml:

plugins:
  github:
    enabled: true
    username: "your-username"
  spotify:
    enabled: true
    client_id_env: "SPOTIFY_CLIENT_ID"
    client_secret_env: "SPOTIFY_CLIENT_SECRET"
    refresh_token_env: "SPOTIFY_REFRESH_TOKEN"

Community plugins are installed from npm (mcp-me-plugin-*) and auto-discovered. See the Plugin Creation Guide.

Generators vs Plugins

CLI Reference

# Auto-generate profile from multiple data sources
mcp-me generate [--directory] --github <user> [--devto <user>] [--stackoverflow <id>] ...

# Initialize with blank YAML templates (defaults to ~/.mcp-me)
mcp-me init [directory]

# Validate profile YAML files
mcp-me validate [directory]

# Start the MCP server (defaults to ~/.mcp-me)
mcp-me serve [directory]

# Scaffold a new generator or plugin (for contributors)
mcp-me create generator <name> [--category <category>]
mcp-me create plugin <name>

Profile location: Defaults to ~/.mcp-me. Override with MCP_ME_PROFILE_DIR or pass an explicit [directory] argument.

Development

# Clone the repo
git clone https://github.com/paladini/mcp-me.git
cd mcp-me

# Install dependencies
npm install

# Run tests
npm test

# Build
npm run build

# Run in dev mode
npm run dev

FAQ

What is MCP (Model Context Protocol)? MCP is an open standard by Anthropic that connects AI assistants to external data sources. mcp-me uses MCP to let AI assistants read your personal profile data.

Do I need API keys? Most generators use public APIs with no auth needed. Some (Strava, Spotify) require tokens — see the .mcp-me.yaml comments for details.

Can I use this with Claude Desktop / VS Code / Cursor / Windsurf? Yes! Any MCP-compatible AI assistant works. See the Configure Your AI Assistant section.

How do I make AI agents use my profile automatically? Drop an AGENTS.md (for Windsurf, Cline, Codex) or .github/copilot-instructions.md (for GitHub Copilot) into your project. Templates are included — see the Agent Instructions Guide.

How do I add a new data source? Run mcp-me create generator myservice to scaffold a new generator, or see the Generator Creation Guide.

Is my data stored anywhere? No. All data stays local in your YAML files. The MCP server reads from disk — nothing is sent to any cloud.

How many generators are there? Currently 329 registered generators, implemented across 44 generator source files plus 15 batch files.

Contributing

We welcome contributions! Whether it's a new plugin, a bug fix, or documentation improvements — see CONTRIBUTING.md for guidelines.

Maintainers: see Publishing Guide for release and distribution instructions.

License

MIT