talentme-mcp

A Local-First Personal Career Memory & ML Interview Coach for AI Assistants

Blog | Documentation | Roadmap | Community Discussion

🌟 What is TalentMe MCP?

TalentMe MCP is a Local-First professional learning memory system designed for Machine Learning and Software Engineering interview preparation. Acting as a bridge between your local knowledge base and your AI assistant (Cursor, Claude Desktop, Trae, Antigravity, Kiro, Codex), it implements the LLM Wiki pattern to track your learning journey, automate mock interviews, and structure your career growth. Learn more and manage your account at TalentMe.

With TalentMe, your AI agent gains long-term context about your projects, skills, and areas of growth, enabling hyper-personalized coaching without leaking your notes to the public internet.

🚀 Key Features

• Local-First Privacy: Your notes, practice logs, and templates are stored locally in Markdown and an Obsidian-compatible structure.
• Dual-Source Intelligence: Connects to the TalentMe Cloud Platform to dynamically fetch specialized interview prompt steering files and expert evaluation workflows.
• Structured Vault & SQLite Engine: Combines standard folders (concepts/, journal/, resumes/) with a local SQLite database for spaced repetition (Ebbinghaus curve) and mastery tracking.
• Zero-Sync Fast Startup: Starting or refreshing the MCP server is 100% local and instantaneous, bypassing cloud requests to prevent delays.
• Auto IDE Registration: One-click configuration registers the server with Cursor, Claude Desktop, Trae, Antigravity, Kiro, and Codex.

🛠️ CLI Commands

TalentMe comes with an intuitive command-line interface:

1. talentme setup

Your initial onboarding assistant.

• Guides you to choose where to save your learning memory folder.
• Connects your account details (Email, License Key).
• Configures AI write mode behaviors (auto, semi-auto, or manual).
• Configures and registers the MCP server inside your IDEs automatically.

2. talentme start

Starts the background Model Context Protocol (MCP) server.

• Automatic: You rarely need to run this manually; your IDE launches it in the background when it boots.
• Fast: Uses a local skeleton builder if a path is empty, completely avoiding blocking network calls.

3. talentme update

Upgrades the software package to the latest version.

• Pulls latest code updates and re-installs dependencies in place.
• Offers an optional, interactive cloud template synchronization loop.

4. talentme sync

Explicitly syncs templates and professional skills from the cloud on-demand.

📂 Vault Structure

TalentMe initializes an Obsidian-compatible local memory folder with the following layout:

my_memory/
├── concepts/             # Atomic knowledge points (e.g., Transformers, PyTorch, Caching)
├── journal/              # Daily practice notes and study logs
├── projects/             # Deep dives into your past system implementations
├── resumes/              # LaTeX/PDF/Markdown resumes and version histories
├── roles/                # Job descriptions and match reports
├── plans/                # Spaced-repetition study paths and 14-day sprint plans
├── .skills/              # Local protocol prompting rules (e.g., llm-wiki, mock-interview)
├── template.json         # Bootstrapping metadata tracking
└── memory.db             # SQLite engine tracking knowledge decay and mastery levels

🎬 Getting Started

Step 1: Install the Package

Install talentme-mcp into your preferred Python environment (we recommend creating a dedicated virtual environment):

python3 -m venv ~/.talentme_venv
source ~/.talentme_venv/bin/activate
pip install git+https://github.com/airsota/talentme-mcp.git

Step 2: Run Setup

Initialize your memory directory and configure your IDEs:

talentme setup

Follow the interactive prompt to set your memory directory, email, license key, and behavior configurations.

[!TIP] Need a License Key? Sign up on the TalentMe official website to generate your key, access premium knowledge bases, and activate your personal career vault.

Step 3: Use with Your AI Assistant

Open your memory directory in Cursor/Claude/Antigravity and start building you own knowledge base!

💬 Example Prompt: "Review my notes on Transformer self-attention. Ask me 3 hard questions about computational complexity, then log my responses to today's journal."

⚙️ Advanced Integration

If the automatic IDE registration is skipped, you can manually register the MCP server by adding this JSON snippet to your IDE's MCP settings:

{
  "mcpServers": {
    "talentme": {
      "command": "/Users/YOUR_USERNAME/.talentme_venv/bin/talentme",
      "args": ["start"]
    }
  }
}

Configuration Locations

• Claude Code: ~/.claude.json
• Cursor: ~/Library/Application Support/Cursor/User/globalStorage/mcpServers.json
• Claude Desktop: ~/Library/Application Support/Claude/claude_desktop_config.json
• Antigravity / Gemini IDE: ~/.gemini/config/mcp_config.json

📄 License

MIT License. See LICENSE for details.