Detection Mode lets you choose the performance profile for each code review. 4 modes trade bug detection, false positives, speed, and cost against each other — so a dependency bump and a payments migration stop costing you the same thing.
$100 in free usage. Usage-based pricing, no seat fees.
Bug detection, precision, latency, and price for every mode. Balanced is the default; Budget and Ultra sit at either end of the curve.
Trusted by engineering teams
Recall, precision, speed, and cost pull against each other. No single configuration maximizes all four, so Macroscope makes the tradeoff explicit instead of picking for you.
Broad detection surfaces more real bugs and some lower-stakes ones. Tighter detection reports only high-confidence findings. Which you want depends on who reads the comments.
Reviews bill per KB of diff, and the rate is set by the mode. Running everything at maximum depth means paying frontier prices to review a lockfile update.
More parallel agents on stronger models take longer. For routine changes, a quick pass that lands before the author context-switches is worth more than a thorough one that lands after.
Mode is resolved for each individual review, so the same repository can run Budget on chores and Ultra on the migration in the same afternoon.
These are not ranked worse to better. Each is built for a different performance profile, at a price that reflects the compute behind it.
Macroscope tops the Martian open-source code review benchmark as the most balanced code reviewer, and ranks #1 on F1 score. Detection Mode is how you move along that curve without giving up the baseline.
Industry leading bug detection. Recommended for teams who use agents to triage PR comments, where the inconvenience of an occasional false positive is minimal.
Everyday application code. This is what you get if you never touch the setting.
macroscope-balancedA 10 KB review costs $0.50 in Balanced.
From an org-wide default down to a single review. More specific settings win over more general ones, and none of them touch your codebase.
The questions teams ask before they change the setting.
Detection Mode is the performance profile Macroscope uses for a code review. There are 4 — Budget, Balanced, Precise, and Ultra — and each one strikes a different balance between bug detection, false positives, speed, and cost. Balanced is the default. The mode is resolved per review, so a single repository can use every mode in the same week.
Start on Balanced and change it where you have a reason to. Move routine, low-risk pull requests down to Budget to cut cost, move code you read comment-by-comment to Precise to cut noise, and reserve Ultra for the changes where a missed bug would be expensive. The modes are not ranked better to worse; each is built for a different profile.
Code review bills per KB of diff reviewed, with a 10 KB floor per review. Budget is $0.025 per KB, Balanced is $0.05 per KB, Precise is $0.06 per KB, Ultra is $0.20 per KB. That is a 8x spread from the lowest-cost mode to the most expensive, so the mode you pick is the single biggest lever on what a review costs.
No. Ultra finds more bugs than Budget, but it also costs more and takes longer, and on a dependency bump it has very little to find. Precise does not out-detect Balanced — it validates harder, so you get fewer low-value findings rather than more findings. Matching the mode to the change is what makes the system work.
Yes, and that is the intended pattern. Set a repository default in Settings → Repos, pin a personal default in Settings → Personal, override a single PR with a macroscope-<mode> label, or request one review in a specific mode by mentioning @macroscope-app in a comment. More specific settings win over more general ones.
Only a conflict if they name different modes. Macroscope resolves each macroscope-<mode> label to a mode and counts the distinct results, so the same mode written twice — a different casing, or an accepted alias such as macroscope-balanced and macroscope-recall — is a single override and the review runs normally. When the labels resolve to two or more distinct modes the request is genuinely ambiguous, so the correctness check is skipped rather than run in a guessed mode. That is scoped to correctness: the approvability check and any Check Run Agents you have enabled still run. Remove the labels naming the mode you did not want, or name the mode explicitly with an @macroscope-app comment.
No. There is no configuration file to commit and nothing to install in your repository. Every mode is a dashboard setting, a GitHub label, or a comment.
No. Detection Mode changes how much compute a review gets, not what Macroscope can read. Every mode works on every repository you connect.
Connect a repository and Macroscope starts reviewing on Balanced. Change the mode whenever you have a reason to.
Get started free$100 in free usage. No card needed.