HackerRank Chakra: the AI Agent Now Running Job Interviews

The newest participant in the job interview is not the candidate, and it is not the hiring manager. According to TechCrunch, HackerRank made HackerRank Chakra generally available to its customers on October 5, 2026, after roughly six months in beta testing. The company says HackerRank Chakra conducted more than 500,000 interviews during that test window, with Snowflake, Snorkel and Capgemini among the early testers.

That volume makes this more than a routine product launch. It is a live, large-scale experiment in what hiring looks like when an AI agent does the interviewing.

An interviewer that watches you work

HackerRank Chakra conducts the interview itself, according to TechCrunch. Candidates receive a task tied to a real code repository and work inside a canvas that includes an embedded AI assistant. As they code, the agent monitors the workflow in real time and asks contextual follow-up questions, pressing on why a particular approach was chosen or how the solution would change under new constraints.

After the session, hiring teams receive both a score and a narrative of the candidate's reasoning, according to Oossa. The format is designed to collapse the traditional three-stage hiring funnel, which is the recruiter screen, the take-home test and the live engineering interview, into a single session.

"AI fluency" becomes the metric that matters

HackerRank's argument is that the old thing interviews measured has lost its signal. Co-founder and CEO Vivek Ravisankar told TechCrunch that the previous evaluation model graded the output, but since anyone can now produce a working artifact, employers need to see the thinking and judgment behind it. HackerRank Chakra scores what it calls AI fluency, meaning how well a candidate frames a problem for an AI assistant, judges the machine's output, and steers it toward a solution.

The pivot is a strategic reversal for a company that built its name on coding challenges. HackerRank, backed by Y Combinator, has spent more than a decade selling standardized assessments to employers. Its developer network now counts tens of millions of users, and more than 3,000 companies buy its hiring tools, according to Nile1. With HackerRank Chakra, the company is betting that observing the process is worth more than grading the artifact.

The cheating numbers nobody expected

Embedding an AI assistant directly into the interview produced an unexpected security outcome, according to Oossa. HackerRank reports that suspicious-activity flags dropped between 70 and 80 percent during HackerRank Chakra sessions compared with its legacy assessments. The theory, reported by Nile1, is that officially permitting AI use removes the incentive for candidates to covertly consult unauthorized external tools while being monitored. The company stresses that final hiring decisions remain with humans, with the agent providing consistent scoring.

Still, the same numbers raise the questions every automated assessment eventually faces: how the scoring holds up across seniority levels and geographies, and whether consistent scoring is the same thing as fair scoring. HackerRank Chakra will now be tested on exactly that, in the open, by every customer that turns it on.

What to do when your interviewer is an agent

For candidates, the preparation playbook changes shape. Practicing polished solutions matters less than practicing how you reason out loud, because the agent grades the steering, not just the destination. Employers running pilots would do well to measure candidate pass rates, time-to-hire and bias signals before rolling the tool out widely, and to require human review on borderline assessments, according to guidance in the AI Agent Store's October 6 industry briefing.

The symmetry is worth sitting with. Voice agents already screen candidates, and candidates already use AI tools to prepare, sometimes without employers knowing, according to TechCrunch. Now agents are conducting the interview and grading how humans use agents. The interview has become a closed loop of agents evaluating people evaluating agents, which makes the fairness question more urgent, not less.

This is the same direction the broader agent economy is moving: software that once recorded outcomes now observes process. Earlier coverage of Constructor's agentic checkout rollout showed the purchasing side of this trend, with agents trusted to spend money on their owners' behalf. The AI news beat keeps filling with stories about agents acting, not just answering. HackerRank Chakra is the hiring chapter of that story, and it arrived with half a million rehearsals already logged.