MergentMergent

Mergent

Intelligent Code Review Assistant

Mergent reviews your pull requests end-to-end, so your team spends less time on routine work and more time building great software.

Get Started

How Mergent Works

Watch how Mergent transforms a raw Pull Request into a perfect merge candidate.

Step 01

Creating the Pull Request

A developer opens a new Pull Request. Mergent picks it up automatically and kicks off the full review pipeline.

Step 02

PR-Issue Linkage

Mergent uses a live vector database and AI-powered semantic similarity to find and link the most relevant GitHub Issue — even if the developer forgot to reference one. It combines text similarity, date proximity and assignee matching into a multi-signal score.[3]

Step 03

Alignment Validation

After linking a PR to its issue, Mergent analyzes whether the code changes actually implement what the issue asks for. It classifies alignment as Exact, Missing, Tangling, or both — catching scope creep and incomplete implementations before review begins.[4]

Step 04

AI Review Comments

Mergent performs a graph-enhanced code review, catching correctness bugs, security issues, regressions, and broken error handling. Every finding goes through two-pass verification to minimize false positives, and results are posted as inline GitHub review comments with severity levels.[1]

Step 05

PR Enrichment

Every PR receives a comprehensive risk analysis with letter grades (A–E) across nine metrics — from bugs and vulnerabilities to impact size and historically buggy files — plus an interactive dependency graph showing the blast radius of changes.[5]

Step 06

Vibe Reviewing

A Chrome extension that brings an interactive AI review assistant directly onto GitHub PR pages. Chat with AI about the code, draft review comments collaboratively, and verify comment quality — all without leaving the diff view.

Your PR is
ready to merge

Full Visibility Into Your Review Pipeline

Track costs, durations, and token usage per pipeline phase. Toggle features on or off, customize quality rubrics, manage integrations, and monitor 30-day trends — all from one place.

Why Mergent

More Than Just AI Comments

Other code review agents stop at posting suggestions. Mergent covers the entire review lifecycle — from linking issues to coaching reviewers — with deep integrations, a custom context engine, and full visibility through a data dashboard.

End-to-End Workflow

Issue linkage, alignment, enrichment, AI review, and Vibe Reviewing — every stage of the review, not just comments.

  • Six pipeline phases working together
  • Full PR lifecycle, not just comments
  • Configurable per repository

Context Engine

A live graph of every symbol and import chain — agents understand the blast radius of each change.

  • Maps symbols & dependency chains
  • Knows what a small change can break
  • No tokens wasted on irrelevant code

Human-in-the-Loop

Vibe Reviewing — a Chrome extension that adds an interactive AI assistant on every GitHub PR.

  • Chat on the diff with code context
  • Draft comments in one click
  • Persistent threads per PR

Review Quality Check

Every comment passes an AI gate that checks accuracy, tone, and clarity against your rubric.[2]

  • Rejects, revises, or approves
  • Custom rubric per repository
  • Right from the GitHub comment box

Highly Configurable

Toggle phases, tune thresholds, and edit rubrics — per repository, per team.

  • Toggle pipeline phases on / off
  • Adjust severity & risk thresholds
  • Per-repo rubrics and settings

Quick Onboarding

Install the GitHub App, grant access, and Mergent starts reviewing — in minutes, no setup.

  • Install the GitHub App in minutes
  • Automatic codebase indexing
  • First reviews land instantly

Cost Monitoring

Per-phase cost, token usage, and duration for every PR — with trends and org-wide summaries.

  • Per-phase cost & token breakdown
  • 30-day trends across the org
  • Spot expensive phases, optimize budget

Integrations

SonarQube static analysis flows into PR Enrichment risk grades, and Slack search surfaces team discussions in Vibe Reviewing.

  • SonarCloud & self-hosted SonarQube
  • Slack search with permalinks
  • Configurable per workspace

Meet Our Team

Kaan Aydeniz, Developer

Kaan Aydeniz

Developer

Erdem Aydın, Developer

Erdem Aydın

Developer

Oğuzhan Genç, Developer

Oğuzhan Genç

Developer

Furkan Emre Kolukırık, Developer

Furkan Emre Kolukırık

Developer

Hasan Kutluhan Şıpka, Developer

Hasan Kutluhan Şıpka

Developer

Eray Tüzün, Supervisor

Eray Tüzün

Supervisor

Beyza Ermiş, Innovation Expert · Cohere

Beyza Ermiş

Innovation Expert · Cohere

Mert Kalender, Innovation Expert · Google

Mert Kalender

Innovation Expert · Google

References

The research behind Mergent's pipeline.

  1. [1]
    U. Cihan et al., “Automated Code Review in Practice” 2025 IEEE/ACM 47th International Conference on Software Engineering: Software Engineering in Practice (ICSE-SEIP), Ottawa, ON, Canada, 2025, pp. 425-436. doi:10.1109/ICSE-SEIP66354.2025.00043
  2. [2]
    C. Liu, H. Y. Lin and P. Thongtanunam, “Too Noisy To Learn: Enhancing Data Quality for Code Review Comment Generation” 2025 IEEE/ACM 22nd International Conference on Mining Software Repositories (MSR), Ottawa, ON, Canada, 2025, pp. 236-248. doi:10.1109/MSR66628.2025.00043
  3. [3]
    A. Yaşa et al., “Evaluating ReLink for Traceability Link Recovery in Practice” 2025 IEEE International Conference on Software Analysis, Evolution and Reengineering (SANER), Montreal, QC, Canada, 2025, pp. 80-90. doi:10.1109/SANER64311.2025.00016
  4. [4]
    A. T. Işık, H. Kübra Çağlar and E. Tüzün, “Enhancing Pull Request Reviews: Leveraging Large Language Models to Detect Inconsistencies Between Issues and Pull Requests” 2025 IEEE/ACM Second International Conference on AI Foundation Models and Software Engineering (Forge), Ottawa, ON, Canada, 2025, pp. 168-178. doi:10.1109/Forge66646.2025.00027
  5. [5]
    I. S. Göçmen, A. S. Cezayir and E. Tüzün, “Enhanced code reviews using pull request based change impact analysis” Empirical Software Engineering 30, 64 (2025). doi:10.1007/s10664-024-10600-2