# interviewer-mcp [Health: Active]

**Category:** 🎓 Education  
**Repository:** https://github.com/girik-chadha/interviewer-mcp  
**GitHub Stars:** 1  
**npm Downloads (last month):** 135  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/interviewer-mcp

## Description
Full technical interview prep: company briefing, concepts, code deep-dive, mock interviews.

## Tools
Capabilities this server exposes over MCP:

- **ingest_repo** — Fetch + index a GitHub repo into a teachable code map
- **list_sections** — Sections in teaching order, with covered status + weakness scores
- **get_code_section** — One section's code with file context
- **mark_covered** — Mark a section learned (only after you explain it back)
- **get_interview_targets** — Probe-worthy code, weakest spots first
- **log_interview_result** — Score an answer; powers cross-session memory
- **set_job_description** — Store JD + company + CV; powers briefing, bootcamp, and gap questions
- **get_progress** — Coverage, history, top weaknesses — the "welcome back" tool

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "interviewer-mcp": {
    "command": "npx",
    "args": ["-y","interviewer-mcp"],
    "env": {
      "GITHUB_TOKEN": "",
      "INTERVIEWER_DATA_DIR": ""
    }
  }
}
```

**Requires environment variables:** `GITHUB_TOKEN`, `INTERVIEWER_DATA_DIR` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What interviewer-mcp MCP server does

The interviewer-mcp MCP server turns a GitHub repository into a structured interview curriculum. It fetches repository content through the GitHub REST API, filters for source files, divides code into sections, and orders those sections from entry points through commonly imported logic and utilities. The resulting map supports both guided teaching and questions grounded in the candidate's actual code.

The workflow covers five related activities: company briefing, concept bootcamp, code deep-dives, mock interviews, and debriefs. A stored job description, company, and CV can connect repository sections to expected skills and highlight areas the codebase does not demonstrate. This makes the interviewer-mcp MCP server a fit for project-focused technical preparation rather than algorithm practice or general behavioral coaching.

## How it works

The interviewer-mcp MCP server exposes eight MCP tools. `ingest_repo` creates the code map; `list_sections` reports the teaching order, coverage state, and weakness scores; and `get_code_section` returns a selected section with file context. A section is marked complete with `mark_covered` only after the candidate explains it back.

For interview practice, `get_interview_targets` identifies code worth probing and puts weaker areas first. `log_interview_result` records whether an answer was strong, okay, or weak, along with a note. Strong answers reduce a section's weakness over time, while weak answers increase it. `get_progress` provides the returning-session view of coverage, history, and top weaknesses. `set_job_description` stores the job description, company, and CV for briefing, bootcamp, and gap-question use.

Sectioning is regex-based for JavaScript and TypeScript, Python, Go, Java, C#, and Kotlin, with chunking as a fallback. Exact AST-based boundaries are not part of the current implementation.

## Setup and configuration

The MCP path runs with Node.js 20 or newer and can be started through `npx`:

```bash
claude mcp add interviewer -- npx -y interviewer-mcp
```

A GitHub token can be supplied for private repositories and a higher GitHub API limit:

```bash
claude mcp add interviewer --env GITHUB_TOKEN=github_pat_... -- npx -y interviewer-mcp
```

GitHub classic and fine-grained tokens are supported; the documented permission is read-only Contents access. Without a token, the server can use the lower unauthenticated request limit described by the project.

The local data directory defaults to `~/.interviewer-mcp/`. Set `INTERVIEWER_DATA_DIR` to change its location. The local store contains the code cache and interview history.

## Tools and capabilities

- Ingest a GitHub repository without cloning it locally.
- Present source code in ordered, teachable sections.
- Track which sections were explained and identify weak spots.
- Generate interview targets from code patterns such as API calls, authentication handling, raw SQL, concurrency, TODOs, and long uncommented functions.
- Associate preparation with a job description, company, and CV.
- Preserve scored interview answers for later sessions.

The tools are intended to be directed by an MCP client or agent. The server supplies repository context and persistence; the interview dialogue itself is conducted through the connected agent.

## Limitations and notes

This project is aimed at explaining and defending project code. It is not presented as a DSA drilling tool, and the README advises skipping it for generic behavioral preparation without a codebase. Repository analysis currently uses language-family regexes and fallback chunking rather than tree-sitter parsing. Multi-repository interviews and spaced-repetition scheduling are listed as roadmap items, not current capabilities.

A Claude Skill is also documented as a separate, no-install option for claude.ai. That route clones the repository per session and relies on conversation search for prior-session recall; the durable local database and persistent coverage model described here belong to the MCP server path.

_Full upstream README: https://allmcps.com/mcp/interviewer-mcp/readme_