CmdRound

About CmdRound

CmdRound is an assessment platform that lets hiring managers evaluate how engineers work with AI coding agents. A candidate receives a link, gets 55 minutes with Claude Code inside a real codebase, and completes an industry-standard task. The tool records every prompt, diff decision, and test result, then produces a rule-based score across four questions-no LLM judging involved.

Review

CmdRound tackles a gap that's opened up as AI-assisted coding becomes common: most hiring tests still block agents and rely on isolated puzzles. This platform puts candidates in front of a real codebase with Claude Code active, then measures what they actually do with it. The replay and scoring system give hiring managers a window into a candidate's process, not just the final output.

Key Features

• 55-minute timed assessment with Claude Code in a pre-configured codebase. Candidates install nothing-the whole round runs through a link.
• Full session replay that captures prompts written, diffs approved or rejected, time spent reviewing changes, and when tests failed and passed again.
• Rule-based scoring on four questions: Do they verify the AI's work? Does the work hold up? How do they recover when it breaks? How do they direct the agent?
• No LLM judge involved in scoring. The evaluation runs on deterministic rules rather than an AI interpreting candidate performance.
• Demo round available on the website so hiring managers can experience the assessment themselves before sending it to candidates.

Pricing and Value

CmdRound costs $49 for 10 candidate assessments, paid once and valid for a year. The maker hasn't announced other pricing tiers. At roughly $4.90 per candidate, the one-time fee structure differs from per-seat subscription models common in hiring tools. Candidates pay nothing.

Pros

• Scoring uses fixed rules instead of an LLM judge, which makes results repeatable and easier to audit.
• The replay shows not just what a candidate did but how they interacted with the agent-review time, rejections, and test state changes.
• Candidates need no setup or installation; the entire assessment runs through a link.
• Four-question score format keeps the output short enough for a hiring manager to act on quickly.
• One-time pricing for a batch of candidates works for teams hiring in cycles rather than continuously.

Cons

• The tool is not well suited for organizations that want to assess candidates in their own proprietary codebases-only the platform's provided codebases and tasks are available at this stage.
• Roadmap items aren't specified publicly, so it's unclear if features like custom tasks or candidate access to replays will ship.
• With only one assessment format (55 minutes, Claude Code), teams looking to test across different agents or time constraints have no alternatives within the platform.

CmdRound fits small to mid-sized engineering teams that already use AI agents in their workflow and want a hiring signal that reflects that reality. The replay data gives interviewers concrete material for follow-up conversations, and the fixed scoring avoids the opacity that comes with LLM-based evaluation. Teams that need to test candidates inside their own repositories or across multiple AI tools will find the current scope limiting.