Don’t just test people.
Understand how they think.

Proctored One is talent intelligence for AI-native hiring. We’re not here to catch cheaters. We help you see how candidates reason, adapt, collaborate with AI, and work in environments that feel like the real job.

The Shift

Hiring still rewards people who fit the test, not the people who fit the work.

When every candidate is judged by one narrow frame, teams miss the minds that solve differently.

01 · the old frame

Most systems begin with suspicion.

Traditional assessments ask every person to prove themselves the same way, then reward the few who adapt to that artificial pattern. That filters people out before you understand what makes them valuable.

02 · the real signal

Output alone hides the mind.

A final answer cannot show judgment, creativity, tradeoffs, persistence, or how someone reacts when the path is unclear. The thinking process is often more valuable than the polished result.

03 · the new standard

AI fluency is part of modern work.

Using AI should not automatically be treated as cheating. The stronger question is whether candidates use intelligent tools with taste, verification, ownership, and clear reasoning.

Our position

AI use is not the problem.
Unexplained work is.

Great hiring is not about forcing everyone into the same narrow test. It is about understanding how each person thinks, decides, learns, collaborates, and turns tools into outcomes.

In production, AI assistants are part of the ecosystem. Cursor, Claude Code, Copilot, Windsurf, Cody. They make strong people faster and expose how they think: what they ask, what they accept, what they reject, and what they choose to own.

The point of an assessment is not to ban tools or trap candidates. It is to surface the human underneath the workflow. Can they reason about a system? Choose the right abstraction? Recover from ambiguity? Explain why they did what they did?

So we don’t reduce people to a suspicion score. We attribute and contextualize. Every line, prompt, decision, and revision becomes evidence of how someone works, so you can evaluate capability, judgment, and role fit with confidence.

“Code is just the residue of thinking. We hire for the thinking.”
Detection coverage

Every AI coding assistant currently in the wild, detected.

Identified by process signature, network fingerprint, IDE extension footprint, completion-API call patterns, and behavioral signature. Updated weekly as new tools ship.

Cursor
IDE fork
Claude Code
CLI agent
GitHub Copilot
extension
Windsurf
IDE fork
Cody
extension
Codeium
extension
Tabnine
extension
Continue
extension
Aider
CLI agent
Cline
extension
Devin
browser
Replit Agent
browser
Cluely
overlay
Ghost
overlay
UltraCode
extension
Prakrit AI
browser
Amazon Q
extension
+ 23 more
see registry
detected & classifiedadded this month
The Signal Stack

Eight signal layers. One human-readable work story.

Proctored One captures the context around real work: environment, decisions, tools, AI collaboration, and revisions. The result is not a gotcha log. It is a structured explanation of how the work came together.

01 · SYSTEM

Workspace events

Process launches, file changes, clipboard transitions, and environment activity captured as context for how work actually happened.

02 · CAMERA

Identity & presence

Identity and presence checks help verify session integrity without turning every movement into a verdict.

03 · SCREEN

Screen & workflow

Screen recording, display changes, app switching, and workflow transitions show the path behind the final submission.

04 · BEHAVIOR

Decision patterns

Keystroke cadence, paste bursts, focus shifts, and iteration loops reveal how the candidate explores, corrects, and commits.

05 · ACCESS

Environment policies

Define the tools the role actually allows: Cursor, VS Code, terminal, browsers, docs, AI assistants, or custom constraints.

06 · AI USAGE

Human-AI collaboration

Every prompt, accept, reject, edit, and verification step is connected to the work it influenced.

07 · NETWORK

Network & tool use

DNS allowlists, external services, documentation lookups, and network activity show what resources shaped the workflow.

08 · AUDIT

Evidence trail

A tamper-evident, export-ready chronology turns the session into reviewable evidence, not vague suspicion.

The Platform

Turn realistic work into talent intelligence.

01 · Create Assessment

Mirror the role. Define what good looks like.

Link a GitHub repo or choose a realistic template. Set the approved tools, workflows, AI policy, constraints, and rubric that match how your team actually works.

proctored.one / create-assessment
Assessments
All Assessments
Drafts3
Templates
Question Bank
Analytics
Overview
Reports
AI Agent Performance
Settings
Team
Integrations
Billing
Enterprise Plan
15 assessments remaining
AM
Alex Morgan
alex@acmecorp.com
New Assessment
Autosaved
Preview
Create Assessment
1
Setup
2
Rubric
3
Agents
4
Invite
Repository
Connect a GitHub repository
github.com/acme/billing-service
Change Repository
Branch
main
Path (optional)
/
Visibility
Private
Evaluation Rubric
Define custom evaluation criteria
Edit Rubric
Code QualityArchitectureTest CoverageAI Prompt DisciplineError Handling+ 5 more
AI Review Agents (10+)
Select AI agents that will evaluate
Manage Agents
Code Quality
Agent
Architecture
Agent
Test Coverage
Agent
AI Prompt Discipline
Agent
Error Handling
Agent
Security
Agent
Performance
Agent
Readability
Agent
Best Practices
Agent
Documentation
Agent
+ Add Agent
Invite Candidates
Share the assessment link
Copy Link
Secure link·No login required·Expires in 30 days
02 · Candidate Experience

Candidates work like they would on the job.

A browser-based IDE with approved AI tools, terminal, repo context, and real tasks. They build, debug, research, and decide in conditions that reveal their natural working style.

PROCTORED ONE
acme / billing-servicemain
02:48:31TIME REMAINING
AM
EXPLORER
BILLING-SERVICE
src
controllers
invoice.controller.ts
payment.controller.ts
services
invoice.service.ts
payment.service.ts
models
invoice.model.ts
payment.model.ts
utils
logger.ts
validation.ts
routes
invoice.routes.ts
payment.routes.ts
invoice.service.ts×
srcservicesinvoice.service.tsInvoiceService◆ retryPayment
1import { Payment, Invoice } from '../models';
2import { Logger } from '../utils/logger';
3
4const logger = new Logger('InvoiceService')
5
6export class InvoiceService {
7 async retryPayment(invoiceId: string) {
8 const invoice = await Invoice.findById(invoiceId);
9 if (!invoice) {
10 throw new Error('Invoice not found');
11 }
12
13 if (invoice.status === 'paid') {
14 return invoice;
15 }
16
17 try {
18 const payment = await Payment.process(
19 invoice.amount, invoice.customerId);
20 invoice.status = 'paid';
21 invoice.paidAt = new Date();
22 invoice.paymentId = payment.id;
23
24 await invoice.save();
25 logger.info(`Payment retried for ${invoiceId}`);
26 return invoice;
27 } catch (error) {
28 logger.error('Payment retry failed', error);
29 throw new Error('Payment retry failed.');
30 }
31 }
32}
TERMINALPROBLEMS2OUTPUTDEBUG CONSOLE
bash ∨
$ npm run dev
> billing-service@1.0.0 dev
> ts-node-dev --respawn src/index.ts
[INFO] 10:24:15 ts-node-dev ver. 2.0.0 (using ts-node 10.9.1, typescript ver. 5.3.3)
[INFO] 10:24:16 Starting server on http://localhost:3000
CLAUDE CODEBETA
Help me implement exponential backoff for payment retry to handle transient failures.

I'll help you implement exponential backoff for payment retries. Here's how we can update the retry logic:

Suggested changesTS
17 try {
18 + const payment = await Payment.process(invoice.amount,
19 + let lastError: Error | null = null;
20 + const maxRetries = 3;
21 + const delay = 1000; // 1 second
22
23 + for (let attempt = 1; attempt <= maxRetries; attempt++)
24 + const payment = await Payment.process(invoice.amount,
25 + break;
26 + } catch (error) {
27 + lastError = error as Error;

This implements exponential backoff with retries:

  • 1st retry: 1 second delay
  • 2nd retry: 2 seconds delay
  • 3rd retry: 4 seconds delay

Let me know if you'd like me to apply these changes.

Using exponential backoff best practices ∨
Ask Claude to help…
main○ 2◇ 0
Ln 13, Col 1Spaces: 2UTF-8LFTypeScript
03 · Review Results

Review reasoning, workflow, and fit.

Multiple AI agents analyze the submission against your criteria. See the final output, the path taken, where AI helped, what the candidate understood, and whether their workflow aligns with your team.

Open a sample report →
ASSESSMENT
Overview
Candidates
Submissions24
Rubric
Settings
ANALYTICS
Results
AI Agents
Benchmarks
Enterprise Plan
15 assessments remaining · Renews in 12 days
AM
Alex Morgan
alex@acmecorp.com
Assessments Senior Full-Stack Submission #78291
MC
Maya Chen
maya.chen@example.com
2h 48m2h ago
Overall Score
8.4 / 10
Strong Hire
Attribution
61% Human
39% AI
Verdict
Recommend advancing
Strong problem solving, clean code, and solid system design understanding.
AI Rubric Scores
CRITERIASCOREWTWS
Code Quality
8.8
20%1.76
Architecture
8.2
20%1.64
Test Coverage
7.6
20%1.52
AI Prompt Disc.
9.1
15%1.37
Error Handling
8.0
15%1.20
Best Practices
8.3
10%0.83
Total83.2%8.32
Code Attribution Timeline
Human AI
00:0042m1h 24m2h 06m2h 48m
Code Review
Replay
AI Usage
Debrief
FILES
src
controllers
invoice.controller.ts
services
invoice.service.ts
models
invoice.model.ts
utils
logger.ts
services › invoice.service.ts
TypeScript
1
export class InvoiceService {
2
  async retryPayment(id: string) {
3
    const inv = await Invoice.findById(id);
4
    if (!inv) throw new Error('Not found');
5
    if (inv.status === 'paid') return inv;
6
    try {
7
      const p = await Payment.process(inv.amount);
8
      inv.status = 'paid';
9
      await inv.save();
10
      return inv;
11
    } catch (e) {
12
      logger.error('Retry failed', e);
13
      throw e;
14
    }
15
  }
16
}
AI SUGGESTED CHANGE View in Chat+3 more
- await Payment.process(inv.amount);
+ await Payment.process(inv.amount, { retry: 3, backoff: 'exp' });
AI Agent Evaluations
Code Quality
8.8 / 10
Readable, well-named code.
Architecture
8.2 / 10
Clean service/layer separation.
Test Coverage
7.6 / 10
Good coverage, some gaps.
AI Prompt Discipline
9.1 / 10
Minimal AI over-reliance.
Error Handling
8.0 / 10
Meaningful, consistent errors.
+5 Agents
View all
How It Works

From interview loop to work intelligence.

Replace disconnected screens with one realistic session that explains capability, process, and fit.

10+
Specialized AI agents evaluate reasoning, architecture, quality, and collaboration against your criteria.
Every
line
Connected to context: human-written, AI-generated, AI-modified, pasted, revised, and reviewed.
Full
replay
Every prompt, decision, correction, and workflow turn becomes reviewable hiring evidence.
Compare

Move from filtering people out to understanding people clearly.

Other platforms optimize for standardized screening. Proctored One is built for realistic work, human-AI collaboration, and the evidence teams need to recognize different kinds of excellence.

HackerRank
CodeSignal
Rounds.so
Proctored One
What candidates do
Algorithmic tasks in a sandbox
Standardized coding tasks in a sandbox
AI-resistant puzzles + DSA
Real work in an environment that mirrors the role
How AI is treated
Built-in AI copilot (guarded modes)
Cosmo AI copilot (GPT-4o)
AI available, problems designed to limit it
Allowed, attributed, and evaluated as collaboration signal
Process visibility
Full AI transcripts + fluency grading
Full Cosmo conversation log
Limited visibility
Prompts, decisions, edits, rejections, and revisions connected
How work is understood
Pass/fail tests + automated review
Standardized scoring framework
AI-powered evaluation
10+ agents analyze reasoning, quality, judgment, and fit
Human-AI attribution
Not available
Not available
Not available
Every line classified and contextualized by workflow history
Workflow realism
No, sandbox only
No, sandbox only
No, standardized problems only
Your repo, approved tools, realistic constraints, real tasks
Loop impact
Varies, automated and human
Automated + AI / human interviews
Minimal, AI-powered
Async by default, with richer evidence for faster decisions
What you get

Not just a score. A map of how someone thinks and works.

Multi-agent talent intelligence

10+ independent AI agents evaluate reasoning, architecture, quality, judgment, and workflow fit against your rubric.

CQcode-quality.agent8.8
ARarchitecture.agent8.2
TCtest-coverage.agent7.6
APai-prompt.agent9.1
EHerror-handling.agent8.0
SEsecurity.agent7.9

Real environment, real strengths

Candidates work in conditions that mirror the role. Real code, real constraints, real tools, real signal.

AI collaboration signal

See where candidates use AI, how they guide it, what they verify, and where they rely on their own judgment.

Human-AI attribution

Every line connected to its origin and workflow context: human-written, AI-generated, or AI-modified.

15  export function retryInvoice(id: string, attempt = 0) {
16    if (attempt >= MAX_RETRIES) throw new RetryExhausted(id);
17    const delay = backoffMs(attempt);
18    await sleep(delay);
19    return schedule(id, attempt + 1);
20  }
Human AI generated
Pricing

Simple plans that scale with your hiring.

Basic
For teams starting with AI-native talent intelligence.
$199/mo
5 assessments / month
  • Realistic work sessions
  • AI-generated reasoning debriefs
  • Workflow replay
  • Human-AI attribution analysis
Get Started
Enterprise
For orgs with high-volume or custom needs.
Custom
Unlimited assessments
  • Everything in Premium
  • Custom work environment templates
  • SSO & team management
  • Dedicated account manager
  • Custom integrations & SLAs
Book a Call
Need more sessions? $49 per additional session. Includes $5 Claude Code budget per session.
FAQ

Common questions

Still have questions about how Proctored One helps your team understand candidates?

Book a Call
Is Proctored One a cheating-detection tool?+
No. It captures evidence, but the goal is understanding, not suspicion. We help teams see how candidates reason, use AI, make decisions, and work under realistic conditions.
Is our source code secure?+
Each candidate gets an ephemeral, isolated fork of your repo. Sessions are sandboxed, network egress is restricted, and the workspace is destroyed after submission. We never train on your code, and SOC 2 Type II is in progress.
How much does it cost?+
Plans start at $199/mo for 5 assessments. Additional assessments are $49 each. Each assessment includes a $5 Claude Code budget for the candidate.
How do I get started?+
Book a 20-minute demo. We'll walk through a sample report from your repo and have your first assessment live the same day.
Do candidates need experience with AI coding tools?+
No. The IDE includes a short Claude Code primer. We don't penalize unfamiliarity; we evaluate how thoughtfully the tool was used, not how often it appeared.
Do you punish candidates for using AI?+
No. AI usage is context, not an automatic negative signal. Employers define what is allowed for the role, and Proctored One shows whether the candidate used AI with judgment, verification, and ownership.
What languages and frameworks are supported?+
Anything that builds on a standard Linux container: TS/JS, Python, Go, Rust, Ruby, Java, C#, and more. If it has a Dockerfile, we run it.
How do candidates feel about the experience?+
Candidates consistently say it feels closer to the actual job than a puzzle, take-home, or timed Zoom round because they can show how they really think and work.
Can I see a sample report before committing?+
Yes. Open the demo control panel for a full sample submission with multi-agent scoring, attribution, and replay.
Does Proctored One integrate with our ATS?+
Greenhouse, Ashby, and Lever are first-class. Anything else, we ship a webhook in a day.
Understand your best hire

See the people behind the output.

Stop guessing from polished answers. Understand who reasons deeply, uses AI with judgment, adapts under realistic conditions, and naturally fits the way your team works.

Book a DemoOpen the Control Panel →