Code Scans automates codebase investigations

Tell Devin what you want to achieve, and it investigates your codebase, evaluates findings, and turns them into pull requests ready to review.

Many goals stay on the backlog

Engineering goals like improving SEO or reducing maintenance overhead often stall because investigating where to start is a massive project, and then implementing all the changes takes even longer.

Agentic MapReduce powers the system

Code Scans uses the same architecture built for Devin Security Swarm, breaking large investigations into focused batches, distributing them across parallel agents, and synthesizing findings into one report.

Early testers see major results

Teams testing Code Scans report a 96% pull request merge rate and estimate saving over 700 engineering hours, with significant improvements in software quality, bug fixes, stability, performance, and compliance.

“Code Scans had a significant impact on the things we care most about: software quality, bug fixes, stability, performance, and compliance, including security and privacy.”

Define your goal, not the files

You don't need to know which files to inspect. Instead, you define what you're looking for, and Devin works with you to establish scope, settings, and what counts as a finding.

Rust compilation time cut by 64%

Code Scans identified opportunities in the Dioxus repository to streamline dependencies and make heavyweight features opt-in, reducing clean debug build time from 58.6 seconds to 21.0 seconds across 22 workspace crates.

SEO health score jumped 12 points

A scan across devin.ai and cognition.com surfaced 44 SEO findings. After fixes, Ahrefs health score increased from 87 to 92 on devin.ai, slow pages decreased by 73%, and missing image alt text was eliminated.

Four-phase investigation process

Code Scans works in four phases: Plan defines rules for relevant code, Shard divides matching code into batches, Map has parallel agents investigate each batch, and Reduce combines findings into one prioritized report.

Scan for performance issues

Code Scans can find slow paths, redundant computation, and missed caching opportunities. It can also identify N+1 database queries, unnecessary round trips, and other inefficient data-access patterns.

Scan for test coverage gaps

Code Scans identifies important flows, edge cases, and failure paths that aren't covered by tests, helping teams prioritize what to test next.

Scan for dead code and cleanup

Find unused functions, modules, dependencies, and obsolete feature flags with evidence for safe removal. Also identify redundant abstractions, excessive boilerplate, and code that can be simplified.

Scan for compliance and accessibility

Code Scans can investigate sensitive-data handling, audit trails, and retention logic against your policies. It also finds missing labels, broken keyboard interactions, and gaps against WCAG criteria.

Start scanning today

Type /scan in your Devin session, specify what you want to investigate, build the scope together, and start optimizing your codebase. Visit devin.ai to try Code Scans.