Starter
$63/ month
Billed annually · $756 / year
- Live + async interviews
- AI-enabled coding
- Up to 3 team members
- 25 candidates a month, 5 working at once
- 600 live call minutes a month
Interview engineers the way they work now.
Live coding interviews in a real sandbox. Candidates read an unfamiliar codebase, find the bugs and talk it through, with Claude/Codex in the terminal.
Set up in minutes · EU hosted · Cancel any month
routes.tstodos.test.ts
1router.post('/todos', async (req, res) => {
2 const { title } = req.body;
+ if (!title?.trim()) {
+ return res.status(400).json({ error: 'title required' });
+ }
6 const row = await db.one(
7 'insert into todos (title) values ($1) returning *',
8 [title],
9 );
10 res.status(201).json(row);
11});
Ada · candidateYou · reviewerBefore you accept it: what if title isn’t a string?
TERMINAL · claude
● The route inserts req.body.title without validating it.
● Update(src/routes.ts)
⎿ Added 3 lines
$ node --test
✓ pass 4 · fail 0
Reviewer log · live
00:41Ran node --test · 2 red
01:12Prompt why do the two POST tests fail?
02:20Ran node --test · 4 green
Prompts1
Tests4/4
Questions1
The position
Kendor is the first interview platform where candidates work live with Claude/Codex in the terminal, and you see every prompt, edit and test run.
Where interviews are going
Google is piloting a code comprehension round: read an unfamiliar codebase, find the bugs, talk through the design, with an AI assistant at hand. That is a Kendor interview, on your own repo.
Four steps.
One honest signal.
No question bank, no webcam theatre. Pick a project, let them work, read what happened, advance them.
Pick
A Kendor project or your own repo. Decide if AI is on.
01
Sit
A booted sandbox. Editor, terminal, database and preview.
02
Review
Timeline, counts, playback and an AI summary.
03
Advance
Results write back to the stage in your ATS.
04
The environmentYour repo, running in seconds.Cloned from GitHub into an isolated container with Node, PostgreSQL and Claude/Codex ready. Tests are read-only. Nothing for the candidate to install.
kendor sandbox createReady in 1.2 s
acme/orders-apimain · 14 files
Your repo · GitHub
src
routes.ts
db.ts
schema.sql
tests
todos.test.ts
kendor.yaml
package.json
Read-only · tests/
routes.tsdb.ts
src › routes.ts
1import { Router } from 'express';
2import { db } from './db';
3
4export const router = Router();
5
6router.get('/orders', async (req, res) => {
7 const page = Number(req.query.page ?? 1);
8 const rows = await db.many(
9 'select * from orders limit 20 offset $1',
10 [(page - 1) * 20],
11 );
12 res.json(rows);
13});
mainTypeScriptLn 6, Col 14
TERMINAL · kendor
$ kendor sandbox create
✓ cloned acme/orders-api · 14 files
✓ container ready in 1.2 s
✓ tests/ marked read-only
✓ dev server on :5173
✓ Claude Code on the org key
Running
Node22.23
PostgreSQL18.6
Valkey9.1
Claude/Codex
The reviewOne session. Every move.A real review page, scrubbed. Drag the playhead to see what the candidate did, what Claude/Codex did, and when the tests turned green.
00:38:48 / 00:45:00
Ada Okafor · Senior backend interview
0m5m10m15m20m25m30m35m40m45m
Candidate
Claude
Tests
Test runs4
AI prompts2
Pastes1
Lines by hand32
At this momentAll four green. Submitted with six minutes left.
TipThe paste at 30:00 is worth a question in the debrief.
AI policyYou decide if AI is on.Off, Claude Code or Codex, per assessment. On your own key, with every prompt in the reviewer log.
TERMINAL
$ claude
command not found: AI agents are off for this screen.
$ node --test
✓ 2 passed · 2 failed
Reviewer log
00:41 Ran node --test · 2 red
09:12 routes.ts edited by hand
15:48 Paste · 3 lines
38:50 4 green · submitted
TipOrg keys, budget capped per candidate.
StacksTheir real stack. Not a quiz.Pick a Kendor project or import your own repo. Databases run inside the sandbox, and the tests stay read-only in the repo.
$ kendor new --stack 7 languages · 2 databases · Docker + k8s
HackathonsHost a hackathon. See how it was built.A public page in seven languages, applications without accounts, every decision in one email, and a real sandbox with their own AI for each builder.
kendor.io/hackathons/acme-search-dayEnterpriseHow it works
01Publish
AHosted by AcmeApplications open
Rebuild search in a day
- Sat, Nov 21 · build day9 h on your clock
- OnlineIn your browser: VS Code, a terminal and your AI agent
Applied: 312 from 41 countries
ApplyEN ES FR DE IT PT RU
02Select
Applications87 of 100 places filled
- AOAda OkaforPT · github.com/adaokAccepted
- LBLinus BergSE · github.com/lbergAccepted
- MTMei TanSG · github.com/meitanWaitlisted
- OHOmar HaddadJO · github.com/ohaddadTo review
Not emailed yet: 24Email decisions
03Build and review
Building nowLive: 86
06:12:48left on their own clock
$ claude
✓ Signed in with their own account
> make search tolerate typos
- Test runs
- 23
- AI prompts
- 41
- Tests
- 12 / 12
When the clock runs out, the work is saved and submitted, ready for review.
PricingSimple. Cancel any month.Prices in USD, billed monthly or yearly. No per-candidate fees.
ProPopular
$400/ month
Billed annually · $4,800 / year
- Unlimited team members and call minutes
- Live + async interviews
- Your own repositories
- AI-enabled coding
- ATS integrations (Ashby, Greenhouse, Lever)
- Interview review and summaries
Enterprise
Custom
- Hackathons for hiring
- Advanced hiring analytics
- SSO and access controls
- Audit and retention policies
- Custom integrations
- Procurement and security support
AI agents run on your own Anthropic or OpenAI account; model usage is billed by your provider, not by Kendor. Fair use applies — see our Terms.
TipPro has unlimited seats and call minutes.
Your turn
Hire by the work.
Set up in minutes · 45-minute sessions · EU hosted
Erased by default
Sandboxes and recordings follow your retention window.
EU hosted
Candidate data is stored in the EU.
Your keys
Agents run on your own Anthropic or OpenAI account.
Security review
SSO with SAML and OIDC, audit policies, a public sub-processor list.