AI Agent

Mergent

From bug report to verified fix, without the backlog wait.

RoleSolo build

Year2026

TypeAI Agent

Launch site

Demo

Open on YouTube ↗

Screens

Mergent — Understand the issue — Understand the issue

Mergent — Fix, tests, preview — Fix, tests, preview

Mergent — Built for QA and engineering — Built for QA and engineering

Overview

Mergent turns a tester’s plain-language bug report into a developer-reviewed pull request. The agent reads the report, explores the codebase in an isolated sandbox, drafts an implementation plan and waits for a human to approve it, then writes the fix, runs the tests, and ships it to a live preview URL so the tester can confirm it actually worked — before opening the PR. No agent-framework graph: the tool-calling loop runs on the Vercel AI SDK, with plain TypeScript orchestrating explore → plan → approve → code → preview → PR.

What it does

Bug reports in plain language — no specs or stack traces required
Human-in-the-loop plan approval before any code is written
Isolated sandbox with cached codebase context between runs
A preview deployment on every run for testers to verify the fix
A pull request carrying the plan, the diff and the test output
Follow-up prompts until the issue is actually closed

Built with

Next.js 15React 19TypeScriptTailwind v4Vercel AI SDKAnthropic + OpenAIPostgresDrizzle ORMSupabaseStripeGitHub App OAuthGSAPThree.js