⚡Model Context Protocol (MCP) Official Portal

TalentMe Official MCP Guide & Configuration

Seamlessly bridge your IDE's AI Agent (Cursor / Claude Code / Cline / Antigravity) with TalentMe's cloud knowledge base and local Obsidian second brain via the Model Context Protocol (MCP).

🐈 GitHub Repo (talentme-mcp)Common CLI Commands ➔

Hybrid Cloud & MCP Flow

The cloud repository serves dynamically updated professional knowledge bases and Agent Skills. The local FastMCP server drives your local IDE (Cursor/Claude) offline without uploading your markdown vault.

☁️

Cloud Repository

Skills & Templates

• Live Q&A Updates• Advanced Prompt Library

⚡HTTPS API

💻

Local MCP Client

FastMCP Server

• Local Database (memory.db)• Strict Offline Mode Supported

⚡STDIO MCP

🤖

IDE Integration

Cursor & Claude

• Natural Language Prompting• Context-Aware Responses

🛡️ Your notes stay on your computer, forever.

TalentMe will never read, upload, or cache any of your local notes. The only optional upload is desensitized study statistics (e.g., "Reviewed 5 topics this week") to generate radars and heatmaps. It even works offline—MCP fully supports offline mode.

🔍 Technical Details (Desensitization Flow)▼

To ensure absolute privacy, background auto-uploads are strictly banned. Local stats are synced only via your clicks. The cloud server deletes the uploaded DB file immediately after updating telemetry aggregates.

💻 Local User Environment

📄 markdown_notes/ (.md)
🗄️ memory.db (Logs & Mastery)
🔒 100% Private (Never leaves local machine)

⚡ Local Sync via Port 8002

JSON Aggregates Only

→

☁️ Cloud Web Server (Port 8001)

📥 Receives desensitized JSON
💥 Temp DB deleted immediately post-parsing!
📊 Saved: Aggregates, Radars & Heatmaps

Diagnostic Dashboard Prototypes

High-fidelity diagnostic metrics help you track preparation gaps. Includes knowledge radars, GitHub contribution heatmap matrix, and double-linked node graph topology.

🕸️ Mastery Radar Preview

Average Mastery score: 4.2 / 5.0

📅 Activity Heatmap Preview

Consistent prep active check indicator.

🕸️ Double-Link Knowledge Graph

Driven by local private Obsidian bidirectional graph.

TalentMe MCP Local-Cloud Bi-directional Toolchain Architecture

💻Your AI IDE (Cursor / Claude Code / Cline / Antigravity)

Intent Recognition & Tool Calling

↓

🤖TalentMe MCP Server (talentme-mcp)20+ Built-in System Tools with Dual Memory Sync

🗂️

🗂️ Local Wiki

Build structured notes, scan for dead links, and maintain wiki health.

☁️

☁️ Cloud Mastery Graph

Sync real-time progress, turning your coding history into structured data.

👨‍🏫

👨‍🏫 Expert Coaching

Manage mock interview states and import expert rubrics for precise guidance.

🚀

Method 1: One-Click Installer Script (Recommended Bash)

We provide an automated installer script that handles environment validation, creates an isolated Python virtualenv to avoid polluting global packages, and guides you through database and IDE (Cursor/Claude Desktop) pairing.

Execute this command in your macOS / Linux terminal:

curl -fsSL https://talentme.airsota.com/install.sh | bash

The installer script will automatically:

Verify python3 >= 3.10 and git dependencies.
Install and compile the client inside a secure sandbox (~/.talentme_venv).
Prompt you to enter your registered email and License Key.
Automatically update Cursor and Claude Desktop config with absolute paths.

💡 Windows or Manual PIP Install: On Windows or non-curl systems, you can directly run: pip install talentme-cli && talentme-cli setup

⚙️

Method 2: Manual IDE Configuration (Alternative JSON)

If you are using extensions like Cline or prefer to manually control your config files, please refer to the guide below.

Step 1: Locate Config File

Open VSCode, find the MCP config option in the Cline extension settings, or directly edit the cline_mcp_settings.json file in your workspace root.

Step 2: Enter JSON Configuration

Paste the following configuration into the config file. Remember to replace YOUR_LICENSE_KEY with your TalentMe subscription key.

{
  "mcpServers": {
    "talentme": {
      "command": "talentme",
      "args": [
        "start"
      ],
      "env": {
        "TALENTME_EMAIL": "YOUR_ACCOUNT_EMAIL",
        "TALENTME_LICENSE_KEY": "YOUR_LICENSE_KEY",
        "TALENTME_MEMORY_PATH": "YOUR_LOCAL_MEMORY_PATH"
      }
    }
  }
}

💡 Note: If your IDE shows 'spawn talentme ENOENT' (Command not found), this is because GUI apps do not inherit terminal paths. Please replace 'command' with the absolute path on your system, e.g., '~/.talentme_venv/bin/talentme' on macOS/Linux or 'C:\Users\YOUR_NAME\.talentme_venv\Scripts\talentme.exe' on Windows.

Step 3: Test Connection

Save the configuration and restart your IDE/client. Upon successful connection, you will see a hammer 🛠️ icon above the Agent's prompt input box. Click it to see all available Tools provided by talentme.

1-Click Config to Supercharge Your Agent

Transform local LLMs and private notes into your personal AI career assistant in 1 minute

🧭 The TalentMe Journey

💻

1. Quick Installation

1-click local setup, seamless Claude/Cursor integration

☁️

2. Cloud Sync

Connect License, sync massive question banks & experiences

📝

3. Daily Learning

Agent tracks your weak points automatically while you use it

📊

4. Dashboard Analytics

View knowledge graphs & learning trajectories on the web

🎓

5. Expert Mock

Share snapshot instantly for targeted mock interviews with human experts

Claude Desktop / Cursor IDE

⚡

Ask your Agent...

🧠 Exclusive AI Q&A (Query)

Break down knowledge silos. In your favorite coding environment, the AI simultaneously queries your "Local Private Notes" and our "Cloud High-Frequency DB" to provide the perfect tailored answer.

🗄️ Local Context (Your Notes)

📄markdown_notes/attention.md

Matches: Q, K, V definitions

☁️ Cloud DB (Live Interview Qs)

⚡TalentMe Live Repo

Matches: 2026 SV Interview Trends

▼

IDE Chat (Cursor / Windsurf)

👤

Explain the QKV matrix calculation in Transformers. Please combine my local notes with the latest tech-giant interview trends.

🤖

🤖 TalentMe Agent

Based on your local note markdown_notes/attention.md, you've summarized that Q, K, and V stand for Query, Key, and Value.

Merging this with our Cloud DB's '2026 Silicon Valley High-Frequency Bank', interviewers often ask a follow-up:

Why do we divide by √d_k before applying Softmax?
Because when the dimension d_k is large, the dot products grow large in magnitude, pushing the softmax function into regions where it has extremely small gradients (vanishing gradients). Scaling by √d_k keeps the variance close to 1, stabilizing the gradients.

Tip: I highly recommend adding this detail to your local note.

⚡ MCP Configured? Try these commands in your IDE Chat (Click to copy):

/tm status - Audit my current mastery levels & weak spots

/tm mock google - Start a tailored mock interview for Google MLE

/tm sync - Sync my local markdown notes with the DB

Check my resume... - Check env & compile my tectonic resume PDF

Looking for CLI command reference?

View CLI Tools & Commands List →