⠀⠀⠁⠀⠀
⠀⡴⣿⣦⠀
⢰⣿⣿⣿⢶
KILN
Sign in
TECHNICAL ASSESSMENT · AGENT ERA

See how engineers actually build with AI.

NO MCQs/NO LEETCODE/NO WHITEBOARD

Kiln drops every candidate into a real cloud dev sandbox with a coding agent. They ship something real — then it's graded on 30+ research-backed engineering signals that tell real engineers from code that just compiles.

See how it works
Trusted by teams hiring the next generation of engineers.
Candidate scorecard GRADED
87/100
Overall signal
STRONG HIRE
Problem decomposition8.4
Correctness9.2
Code quality7.6
Agent direction8.8
30+ signals · backed by full transcript
kiln.dev/s/ax9f2c1e
SANDBOX LIVE
Kiln agent RUNNING
CANDIDATE
Add rate limiting to /ingest — 100 req/min per API key, return 429 with a Retry-After header.
KILN
On it. I'll add a token-bucket middleware backed by Redis and wire it into the ingest router. Writing middleware/rateLimit.ts now.
edited rateLimit.ts · +34 −0
Reasoning
rateLimit.ts ingest.ts
1
2
3
4
5
6
7
8
9
// token bucket, 100/min per key
export const rateLimit = async (req, res, next) => {
  const key = req.header("x-api-key");
  const ok = await bucket.take(key, 100);
  if (!ok) return res
    .set("Retry-After", "60")
    .status(429).end();
  next();
};
$ npm test -- rateLimit
✓ limits to 100 req/min per key
✓ returns 429 + Retry-After
3 passing (412ms)
The flow

One assessment, end to end.

≡ × AGENT
» add rate limiting to /ingest
writing rateLimit.ts
$ npm test ✓ 3 passing
» return 429 + Retry-After
edited ingest.ts +12 −2
▸ reasoning
DECOMP CORRECT QUALITY AGENT SPEED
01 · CANDIDATE-DIRECTED AGENT
≡ × KILN SANDBOX
BUILDING
   **
   **
   **
********
   **
   **
   **
   **
   **
   **
********
   **
   **
   **
CPU: 8× · RAM: 4 GB · 02 · LIVE SANDBOX
≡ × REPORT
SCORE
RUBRIC
LOG
download · evidence-backed
03 · GRADED REPORT
How it works

From prompt to graded report in three steps.

01
Create an assessment

Write a realistic task and a rubric of engineering dimensions. Set a time and token budget. Reuse it for every candidate for the role.

02
Candidate builds with the agent

They get a locked-down sandbox and the Kiln agent, right in the browser. They decompose, direct, and ship — exactly how they'd work on the job.

03
Get an LLM-graded report

Every dimension is scored against your rubric and backed by the full candidate-and-agent transcript. Compare candidates on the same scale.

Problem decomposition
Correctness
Code quality
Agent direction
Features

Everything a static coding test can't tell you.

01
Real work, not trivia

Candidates ship in a live sandbox with a coding agent — building, testing, and debugging real code. No whiteboards, no leetcode.

02
Grades the AI-era skill

Measures how well someone directs an agent — scoping, reviewing, and correcting it — not just how much syntax they can recall from memory.

03
Evidence, not vibes

Every score is backed by the rubric and the complete candidate-and-agent transcript. Click any number to read exactly why it was given.

04
Secure & isolated by design

Each candidate runs in a locked-down, non-root, network-isolated sandbox. Nothing leaks between sessions and nothing touches your infrastructure.

05
API-first

An SDK to provision assessments, stream live sessions, and pull graded results programmatically — drop Kiln into your own hiring or education flow.

06
Governed for real hiring

Role-based access, audit logging, data retention and erasure, and bias / adverse-impact reporting built in — ready for a real hiring pipeline.

For developers

Two ways to use Kiln.

Pick whichever fits your team. Both are live in minutes — there's no infrastructure to run and nothing to install.

NO CODE 01 · SEND A LINK
Email candidates a session link

Create an assessment, send the link, and get a graded report when they submit. No setup on your side.

API / SDK 02 · INTEGRATE
Build it into your own platform

Provision sessions and pull graded results straight from your hiring or education product — a few lines with the SDK.

SDKS JSJavaScript TSTypeScript GoGo Python
hire.ts TYPESCRIPT
import { Kiln } from "@kiln/sdk";

const kiln = new Kiln({ apiKey: process.env.KILN_API_KEY });

// 1 · create an assessment from a prompt + rubric
const assessment = await kiln.assessments.create({
  title: "Senior Backend — rate limiting",
  prompt: "Add per-key rate limiting to /ingest…",
  rubric: ["decomposition", "correctness",
           "code_quality", "agent_use", "speed"],
  budget: { minutes: 90 },
});

// 2 · provision a sandboxed session for a candidate
const session = await kiln.sessions.create({
  assessmentId: assessment.id,
  candidate: { email: "candidate@example.com" },
});
console.log(session.url); // share with the candidate

// 3 · pull the graded report when they submit
const report = await kiln.reports.get(session.id);
console.log(report.score, report.dimensions);
Security & trust

Safe to run on real candidates.

ALL SYSTEMS OPERATIONAL
ISOLATION
Non-root, network-isolated
Each session runs unprivileged with no outbound network by default.
TENANCY
No cross-session leakage
Sandboxes are single-use and destroyed on submit — nothing carries over.
ACCESS
RBAC & SSO
Role-based access with SSO so only the right people see candidate data.
AUDIT
Full audit logging
Every action on every assessment and report is logged and exportable.
RETENTION
Retention & erasure
Configurable data-retention windows and one-call erasure for compliance.
FAIRNESS
Bias & adverse-impact
Built-in adverse-impact reporting to keep your process fair and defensible.

Assess how they actually build.

Request access and we'll set up your first agent-driven assessment. Early access is rolling out to teams hiring engineers now.

No spam. We'll only email about access.