Cloudflare runs 130,000 AI code reviews at $1 each using multi-agent factory
IndyDevDan breaks down Cloudflare's CI-native multi-agent code review system, which ran 130,000 reviews across 5,000 codebases for just $1 per merge request in a single month.
Score breakdown
Cloudflare's $1-per-review cost across 130,000 reviews demonstrates that multi-agent orchestration can attack the code review bottleneck — described in the source as a constraint where median wait times are often measured in hours — at a scale and price point that manual review cannot match.
- 01Cloudflare ran 130,000 AI code reviews across 5,000 codebases in a single month at $1 per merge request.
- 02The system uses a custom agent harness built on OpenCode, not an off-the-shelf AI code reviewer.
- 03Up to seven specialized reviewer agents cover areas including security, performance, code quality, documentation, release management, and compliance.
IndyDevDan presents his first agentic engineering tier list, using Cloudflare's CI-native AI code review system as the subject. The headline figure is striking: over a single month, Cloudflare ran 130,000 AI code reviews across 5,000 codebases at a cost of $1 per merge request. The video frames this as a masterclass in "tokenomics" — a concept defined as using tokens to generate value and then arbitraging that value for more than it costs — and awards Cloudflare an S tier rating on that dimension.
The system is built on a custom agent harness on top of OpenCode rather than an off-the-shelf AI code reviewer.
The system is built on a custom agent harness on top of OpenCode rather than an off-the-shelf AI code reviewer. It launches up to seven specialized reviewer agents covering areas such as security, performance, code quality, documentation, release management, and compliance. A coordinator agent sits above them, deduplicating findings, judging results, and posting a single structured review. The video highlights several architectural choices as noteworthy: a plug-in architecture that makes the factory extensible across thousands of repos, JSONL streaming for real-time agent observability and retry logic, and risk-tiered compute that avoids sending the full agent team to review minor changes like typo fixes.
The video also contextualizes the problem Cloudflare is solving: code review is described, in Cloudflare's own words, as "one of the most reliable ways to bottleneck an engineering team," with median wait times for a first review often measured in hours. The video credits a blog post by Ryan Skidmore as the source material and ranks multiple factory dimensions including harness engineering, agent specialization, multi-agent orchestration, context engineering, prompt engineering, model flexibility, system resilience, agent observability, and developer experience.
Key facts
- 01Cloudflare ran 130,000 AI code reviews across 5,000 codebases in a single month at $1 per merge request.
- 02The system uses a custom agent harness built on OpenCode, not an off-the-shelf AI code reviewer.
- 03Up to seven specialized reviewer agents cover areas including security, performance, code quality, documentation, release management, and compliance.
- 04A coordinator agent deduplicates findings, judges results, and posts a single structured review.
- 05JSONL streaming provides real-time agent observability and retry logic.
- 06Risk-tiered compute means minor changes like typo fixes do not trigger the full agent team.
- 07The video awards Cloudflare an S tier rating for tokenomics and credits a blog post by Ryan Skidmore as the source material.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 9, 2026 · 17:05 UTC. How this works →