Interview Coach MCP

An MCP-powered personal interview preparation assistant that helps track coding-problem attempts, mistakes, insights, and progress through natural language interactions with an LLM.

The project uses the Model Context Protocol (MCP) to expose interview preparation capabilities as tools that an LLM can discover and invoke.

---

Why Interview Coach?

While preparing for coding interviews, solving problems is only part of the process. Understanding recurring mistakes, tracking confidence, revisiting weak areas, and learning from previous attempts are equally important.

Interview Coach is designed to turn this preparation history into persistent memory that an AI assistant can use to provide personalized guidance.

Instead of manually interacting with a database, the user can interact with the coach using natural language.

For example:

"I just solved Network Delay Time. It took me 28 minutes. I initially tried BFS before realizing I needed Dijkstra. My confidence is 8/10."

The LLM can interpret this interaction and use the appropriate MCP tool to store the attempt.

---

Architecture

                    User
                     │
                     ▼
              ┌──────────────┐
              │ Claude / LLM │
              └──────┬───────┘
                     │
                 MCP Protocol
                     │
                     ▼
          ┌──────────────────────┐
          │ Interview Coach MCP  │
          │       Server         │
          │                      │
          │  search_problems()   │
          │  log_attempt()       │
          │  hello()             │
          └──────────┬───────────┘
                     │
                 SQLAlchemy
                     │
                     ▼
              ┌─────────────┐
              │    SQLite   │
              │             │
              │  Problems   │
              │  Attempts   │
              │  Notes      │
              └─────────────┘


Current MCP Tools

hello()

Basic connectivity tool used to verify that the MCP server is running.

search_problems(query)

Searches the local problem database by LeetCode problem title or problem ID.

log_attempt(...)

Stores a user's attempt at a problem, including:

Whether the problem was solved
Time taken
Confidence level
Initial approach
Mistakes
Key insights
Final notes              

Problem
   │
   │ 1 → many
   ▼
Attempt
   │
   │ 1 → many
   ▼
Note