> ## Documentation Index > Fetch the complete documentation index at: https://docs.mergeoa.com/llms.txt > Use this file to discover all available pages before exploring further. # Overview: How Merge evaluates engineers > Merge assessments replace algorithm quizzes with real code review work. Learn how assessments are structured and what signals they produce. A Merge assessment puts candidates inside an actual engineering workflow: they read a real codebase, review a pull request, leave comments on bugs, refactors, and security risks, and then iterate as an AI responds to their feedback and opens a revised PR. This loop runs until the session timer expires, producing a rich record of how each candidate thinks and communicates as an engineer — not just whether they can reverse a linked list. ## What makes a Merge assessment different Merge assessments are grounded in real engineering work — reading a codebase, reviewing a PR, and iterating based on feedback — rather than puzzle-solving under pressure. Candidates engage with the kind of material they would encounter on their first week on the job: existing code with real tradeoffs, real ambiguity, and real consequences. The result is signal your team can discuss with confidence, anchored in observable behavior rather than abstract scoring. ## Assessment structure Each assessment follows the same four-step loop, repeated until time expires: 1. **Read the codebase** — The candidate explores the repository to understand the project's structure, patterns, and conventions. 2. **Review the pull request** — The candidate leaves inline comments identifying bugs, refactor opportunities, security risks, and other concerns, just as they would in a real code review. 3. **Receive an AI revision** — Merge's AI processes the candidate's comments and opens a fresh PR revision that responds to their feedback. 4. **Iterate** — The candidate reviews the revised PR, refines earlier comments, and adds new observations. This loop continues until the session ends. ## Configuration options You control three axes when creating an assessment. Together, they let you target the exact bar your team is hiring for. | Option | What it controls | | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Difficulty** | Codebase complexity and expected review depth. Ranges from Intern to Principal — each level calibrates how subtle the issues are and how deep the AI's revisions go. | | **Specialization** | Engineering domain. Options include Frontend, Backend, Infrastructure, Networking, Core Database, Performance, Distributed Systems, Data Pipelines, Security, Platform, Systems, and APIs. The codebase and PR content reflect the chosen domain. | | **Language surface** | Which programming languages appear in the assessment. Restrict to one or two languages for specialist roles, or allow a broader mix for generalist hiring loops. | ## What you receive When a candidate completes a session, Merge generates a scorecard that captures how they performed across the full review loop. The scorecard shows: * **Reasoning quality** — how accurately the candidate identified real issues versus noise * **Prioritization** — whether they led with the most critical problems or buried them in minor nits * **Iteration behavior** — how their feedback evolved across PR revisions, including whether they incorporated responses thoughtfully * **Communication** — the clarity, specificity, and professionalism of their written comments You receive a structured record you can share with your hiring panel, compare across candidates, and use as a foundation for a debrief conversation. Set up a new assessment from the dashboard in minutes. Tune difficulty, specialization, and language surface for any role. Send assessment links directly or through your connected ATS. Understand the output Merge produces after each session.