menu Documentation menu

AI Tools for Interview

Last updated on • Disponible en Français

Using AI Assist

✅ Your account administrator can make AI Assist available in pads by default in the Organization Settings. You can also turn this feature on or off for an individual pad in the Pad Settings by clicking on the settings icon ⚙️ at the bottom left of the screen.

You and your candidates now have access to current large language models to utilize as a tool in your interviews, including:

  • GPT
  • Claude
  • Gemini
  • Llama

To use it, simply click on the AI Assist tab, select your language model, and enter your prompt.

Screenshot of the CoderPad interface showing a Java project open in the editor. The file tree on the left displays folders for src, test, and files like pom.xml and README.md. The main editor is open to the README.md file with setup instructions for running Java on CoderPad. On the right, the AI Assist panel shows a chat between Ken Swearengen and the AI Assistant discussing an example of using Lodash in JavaScript. At the bottom right, a red-outlined area highlights the AI Assist input box with “Add Context,” “Files,” and “Terminals” buttons, along with a dropdown menu showing model options — including GPT-4o mini (selected), GPT-5, Claude Haiku 3.5, Claude Sonnet 4, Gemini 2.5 Pro, and Llama 4 Maverick.

Some ways you can use the integration:

  • Have candidates showcase their prompt engineering skills
  • Review AI answers to technical questions with candidates to assess their critical thinking processes
  • Allow candidates to use AI like they would any other tool to solve your questions
  • Generate test data
  • Create starter code in different programming languages

✅ All pad users’ prompts and AI output will be saved for review in playback mode.

ℹ️ Want more information on how to use the integration? Checkout the demonstration video on YouTube.

AI modes

✅ The AI Assistant will automatically detect the proper context of your question, no matter what mode you’re in.

Screenshot of the CoderPad interface running a Next.js project. The left sidebar shows a file tree (app, api, db, lib, public, and files like page.tsx, layout.tsx, package.json, and README.md). The main editor displays a README titled “Running Next.js on CoderPad” with PostgreSQL database credentials listed below. At the top, a banner says the waiting room is enabled with a button labeled “Open Pad To Everyone.” On the right, the “AI Assist” panel is open with a ChatGPT interface. Two red callout boxes highlight a dropdown menu with options “Ask” (selected) and “Edit,” shown both enlarged and in the input area near the prompt field. The bottom bar includes controls such as Invite, Start Call, Start Transcription, What’s New, Feedback, and a red “End Interview” button.

Ask mode will allow you to prompt the AI assistant to answer your requests, but won’t allow the assistant any write-access to the code.

To allow the AI assistant the ability to change the code (including the file structure), you’ll need to enable Edit mode.

⏰ Coming soon: Plan mode
AI “plan mode” is when the AI first outlines a clear step-by-step approach to a task instead of jumping straight to the answer. This helps ensure better results by making the process more structured, transparent, and easier to review before execution.

Add other AI tools

✅As of now, there is no way to persist Claude’s or Codex’s installation from a question to a pad. So you’ll need to install it for each individual interview pad for which you want to use it.

⚠️You’ll need your own Anthropic or Codex account and API key to get started.

Anthropic’s Claude

To install Claude in the shell you can either use curl (i.e. curl -fsSL https://claude.ai/install.sh | bash), or run the following commands:

  1. mkdir ~/.npm-global
  2. npm config set prefix '~/.npm-global'
  3. echo 'export PATH=$HOME/.npm-global/bin:$PATH' >> ~/.bashrc
  4. echo 'export SHELL=/bin/bash' >> ~/.bashrc
  5. source ~/.bashrc
  6. npm install -g @anthropic-ai/claude-code
  7. sed -i '1s|.*|#!/usr/bin/env node|' $(npm root -g)/@anthropic-ai/claude-code/cli.js
  8. Run claude --version to verify installation.

Once installation is complete, run claude in the command line, and you’ll receive instructions for set-up, including adding your API key.

Screenshot of a CoderPad interface showing a code editor on the left and a terminal output on the right. The editor displays files and directories, including "src," "models," and "static," with files like "User.ts" and "server.ts." The main text describes running a NodeJS Express app with PostgreSQL. The terminal shows a "Welcome to Claude Code" message in ASCII art, with a URL bar and options for server, shell, and console outputs.

You can then get started using Claude to review and edit your code.

A split-screen view of a coding environment in CoderPad. On the left, a file explorer shows directories like "src" with files such as "User.ts" and "server.ts", "static" with "style.css", and other files like "index.html", "package.json", and "README.md". In the center, a text editor displays instructions for running an Express application with PostgreSQL and using curl for API access, with TypeScript support. On the right, the program output section shows shell instructions and a list of testing update todos.

Codex

Just like Claude, run the following commands:

  1. mkdir ~/.npm-global
  2. npm config set prefix '~/.npm-global'
  3. echo 'export PATH=$HOME/.npm-global/bin:$PATH' >> ~/.bashrc
  4. echo 'export SHELL=/bin/bash' >> ~/.bashrc
  5. source ~/.bashrc
  6. npm i -g @openai/codex