Run a vibe coding interview you can review_

In Codility Interview, candidates build with Claude Code or the AI Assistant in a shared VS Code workspace. You watch the session live, then review the AI activity in the candidate report to see how the candidate thinks.

Request a free trial Our team will contact you about trial access.

Senior backend engineer Live 24:18 End interview

‹1›

‹2›

auto_awesome

checklist

edit_note

              limiter.py
              test_limiter.py
            

1
2
3
4
5
6
7
8
9
10
11
12
13
14
import time
from collections import deque


class RateLimiter:
    def __init__(self, limit: int, window: float):
        self.limit = limit
        self.window = window
        self.calls: dict[str, deque] = {}

    def allow(self, key: str) -> bool:
        now = time.monotonic()
        log = self.calls.setdefault(key, deque())
        while log and now – log[0] > self.window:

Terminal Output AMJK2

candidate@workspace:~/rate-limiter$ claude
>Replace the fixed window with a sliding log per key. Keep the existing tests passing.
●Read limiter.py and test_limiter.py
●Update limiter.py +9 −4
●Run pytest 6 passed
>Why a deque and not a list?

AM Candidate

JK Interviewer

mic videocam chat

Trusted by global organizations including GitHub, SpaceX, Tesla, EY, LSEG, SAP, Barclays and Citi.

Codility is a G2 Leader in Technical Skills Screening, rated 4.6 out of 5 from verified customer reviews.

Today’s failing approach: ban AI, or allow it without a record

Challenge 1

AI is banned in the interview

If your engineers build with AI tools, an interview without AI does not reflect the way your team works today.

Challenge 2

AI runs on the candidate’s own machine

When the candidate uses an AI tool in their own IDE and shares the screen, you see the code appear. Your team has no record of the prompts behind the code after the call ends.

Challenge 3

Each candidate needs an AI account

If your team gives each candidate an AI account and a usage budget, every interview adds setup work. Someone on your team creates the account, sets the budget and removes access afterward.

Run the AI coding interview in a shared VS Code workspace

Codility Interview is a live technical interview platform for engineering teams. The candidate and your interviewer work in the same VS Code workspace, and Codility provides the AI tools inside it. Your team brings its own rubric and either its own task or one from the Codility task library.

Turn AI on or off for each interview

check_circleYour team enables or disables the AI Assistant per assessment. The same setting turns Claude Code on or off.

check_circleAll AI features are off by default at the account level until an admin turns them on.

check_circleYour team can run one interview with AI off and a later interview with AI on.

The difference: You test independent coding and AI-assisted coding in separate interviews.

code_blocks Rate limiter service Same task, 2 interviews

Round 1Alex Morgan

auto_awesomeAI Assistant

terminalClaude Code

def allow(self, key):
    now = time.monotonic()
    log = self.calls[key]
    # candidate writes this
    # part without AI
    

AI Assistant off

Round 2Alex Morgan

auto_awesomeAI Assistant

terminalClaude Code

candidate:~$ claude
> Draft allow() with a sliding log
● Update limiter.py +9
> Window check is off by one.
  Use > not >=
● Update limiter.py +1 −1

AI Assistant on

Terminal Problems Output Alex typing Jordan in terminal

candidate@workspace:~/rate-limiter$ claude

            Claude Code
            cwd: ~/rate-limiter · hosted by Codility
          
>Add a per-key sliding log to RateLimiter and write a test for the window boundary.
● Read limiter.py
● Update limiter.py +9 −4
● Create test_window_boundary +12
● Run pytest 7 passed
>[Jordan] What happens to memory with ten thousand idle keys?
>Evict keys whose newest timestamp is older than the window

JK Jordan Kim · Interviewer

Give candidates Claude Code in the interview terminal

check_circleWhere your account enables Claude Code, the candidate types claude in the workspace terminal and works the task with the agent.

check_circleCodility hosts Claude Code on its own AI infrastructure. The candidate needs no API key, login or AI account.

check_circleYour interviewer and the candidate can both interact with the session while Claude Code runs.

The difference: Your team sets up no AI accounts for candidates.

Review the candidate’s prompts in the candidate report

check_circleDuring the interview, you see each Claude Code prompt and response as it happens.

check_circleCodility saves each Claude Code prompt, edit and correction to the candidate report. Every candidate interaction with the AI Assistant is captured as reviewable AI activity.

check_circleYour interviewers can see how the candidate directed the AI and where the candidate pushed back, alongside the code.

check_circleCodility does not use customer or candidate data to train AI models.

check_circleYour team can also use an AI-Native Task, in which the AI gives some wrong answers on purpose. The candidate report then shows whether the candidate caught the wrong answers.

The difference: Your interviewers base the hiring decision on the code and the prompt record together.

Request a free trial

AM

Alex Morgan Senior backend engineer · Interview · 12 Oct 2026

descriptionCandidate report

Summary Code AI activity Notes

Claude Code prompts7

04:12Prompt

Draft allow() with a fixed window per key.

07:40Correction

A fixed window lets a burst through at the boundary. Switch to a sliding log.

08:05Edit

limiter.py +9 −4

11:32Prompt

Write a test for a request exactly at the window edge.

limiter.pyFinal code

10
11
12
13
14
15
16
17
18
19
def allow(self, key):
    now = time.monotonic()
    log = self.calls.setdefault(
        key, deque())
    while log and \
      now – log[0] > self.window:
        log.popleft()
    if len(log) >= self.limit:
        return False
    log.append(now)

Customer feedback on AI access for candidates

“Everyone’s using AI and it feels sometimes unfair to disregard candidates the access. But then we can monitor it and ensure they’re not using it for the whole thing.”

Engineering leader, consumer analytics

Questions about vibe coding interviews

What is a vibe coding interview?

expand_more

A vibe coding interview is a live coding interview where the candidate builds working software with an AI coding tool. Teams also call it an AI-first interview, an AI coding interview or an AI-assisted coding interview. The interviewer evaluates how the candidate directs the AI, checks the AI’s output and fixes the AI’s mistakes.

Should we allow AI in coding interviews?

expand_more

Yes, if your engineers use AI on the job. Where a role needs coding without AI, your team can turn AI off for that interview.

Which AI tools can candidates use in Codility Interview?

expand_more

The AI Assistant and Claude Code are available to candidates. The AI Assistant helps the candidate clarify the task and will not generate a full solution. Claude Code works as a coding agent in the terminal once an admin turns it on for your account.

How do we see whether Claude Code wrote the whole solution?

expand_more

Your interviewer sees each Claude Code prompt and response as the candidate works. The candidate report keeps each prompt. Your interviewer can then ask the candidate to explain the design decisions behind the code.

Does Codility score the candidate’s use of AI?

expand_more

No. Codility does not grade AI activity automatically. Your interviewers review the AI activity, score the session against your rubric and decide who advances.

Run your next vibe coding interview in Codility

Bring a task from your current interviews and run it with AI turned on.

Request a free trial

"*" indicates required fields

Instagram

This field is for validation purposes and should be left unchanged.

Business email*

First name*

Last name*

Job title*

Company*

Direct phone number

What can we help with? (select all that apply)*

Assess candidates (Screen)

Improve technical interviews (Interview)

Understand my team’s skills (Skills Intelligence)

Replace or evaluate your current platform

I’m an existing customer

Something else

Please tell us more*

Roughly how many hires do you expect this year?*

How many employees does your company have?*

Anything you'd like us to know? (optional)

How did you find us? (optional)

Please tell us how you found us*

utm_source

utm_medium

utm_campaign

utm_content

utm_term

landing_page

referring_page

First touch date

First touch source

First touch medium

First touch campaign

First touch content

First touch term

First touch click ID

First touch referrer

First touch landing page

Most recent touch date

Most recent touch source

Most recent touch medium

Most recent touch campaign

Most recent touch content

Most recent touch term

Most recent touch click ID

Most recent touch referrer

Most recent touch landing page

GCLID

AB variant