State-of-the-art code reviews for less than $2

Written by AI

TLDR: Doistbot runs about 4,500 code reviews a month for Todoist, averaging under $2 each through open models and dynamic routing.
Best for: Engineering teams scaling pull request review with AI
The tip: Start with a focused GitHub Action, then add a review harness, open models, and dynamic routing as volume grows.
What you need: GitHub Actions
A model CLI or review harness
Repository conventions and architecture guidance
Start here: Pass the PR diff and repository guidance to the reviewer
Run it as a pending GitHub review
Track review volume and cost before adding dynamic routing
Filed note: How Todoist built Doistbot from a GitHub Action into a multi-model review harness that handles 4,500 reviews a month for under $2 each.

Highlights: A year later, Doistbot is critical to how we build and ship at Todoist. It runs about 4,500 reviews each month at an average cost below $2 per review.
This is operating at production scale, not as a small experiment.
We went from a scrappy GitHub Action to a proper agentic review harness, then to open models and dynamic routing to keep cost in check.
The system evolved in layers as review volume grew.
Excerpt: By August 2025, frontier models and harnesses were reliable enough to complete complex tasks. Our engineering teams were firing on all cylinders, and PR volume was growing by double-digit percent month over month.
From doist.dev