Greptile's New Pricing
Is Predatory.
Greptile, the YC-backed AI code review startup, quietly switched its pricing in March 2026 from a flat $30 per developer per month to a base-plus-usage model: $30/seat for 50 reviews, then $1 per review thereafter. The change was framed as a response to the rise of AI coding agents. The same change makes Greptile the only major AI code review tool that taxes throughput. And the company's own math, when laid alongside its own claims, does not add up.
01 · THE 10% CLAIM
The Numbers Don't Reconcile.
Greptile's launch post says "less than 10% of active users will exceed the included usage" (v4 announcement). It's the central reassurance.
The CEO's own number contradicts it. On X, Daksh Gupta: "average PR gets 1.2 reviews on greptile" (2026-03-13). At his number, 50 reviews = ~42 PRs per developer per month.
Linear's entire R&D — engineers, designers, PMs, managers — already ships 33 PRs/person/month. Greptile's allowance covers the company-wide median with 9 PRs of headroom. Before engineers are counted separately. Before Greptile's own re-review cycles. Before AI agents enter the picture.
Matt Galligan's actual workflow: 571 PRs in 30 days. The included quota covers 8.8% of it. The bill goes $30 → $500+.
So the "less than 10%" claim only holds if "active user" excludes every developer running AI coding agents — which is the demand curve Greptile is allegedly serving. The 10% who exceed are, by definition, the customers who use the product as marketed.
02 · WHAT CHANGED
The Cap Is Sized For The Pre-Agentic Era.
Four data points on how PR cadence has shifted since 2024. Greptile's 50-review cap was sized against the row above the data; the data is the row below.
PRs/person/month at Linear, March 2026. Doubled from 16in August 2024. Counts the entire R&D org — engineers, designers, PMs, managers.
PRs/day from a single developer running Cursor + Claude Code + Codex in parallel. arlogilbert: "15+ PRs per day of better-than-human code."
of Boris Cherny's contributions to Claude Code over 30 days were written by Claude Code. Anthropic dogfooding the loop end-to-end.
PRs produced by GitHub Copilot agents alone, May 2025 through September 2025. Octoverse stops short of publishing PRs/active-dev/month, but the trend is unambiguous.
Whatever median the 50-cap was sized against, it was a 2024 number. The post-Cursor-Claude-Code-Codex baseline is 2–10× higher and accelerating. Greptile's own framing — "better coding agents have led to a drastic increase in the number of commits reviewed by Greptile" — confirms the shift while pricing as if it hadn't happened.
03 · ALTERNATIVES
The Market Already Priced It Differently.
Side-by-side, dedicated AI code-review tools available as of April 2026. Greptile is the only major review bot that imposes a per-review overage on top of the seat price.
| Tool | Per dev / mo | Review cap | Overage | Notes |
|---|---|---|---|---|
| Greptile Cloud | $30 | 50 reviews | $1 / review | Cap = ~42 PRs/mo at Greptile's own 1.2-reviews-per-PR average. |
| CodeRabbit Pro (annual) | $24 | Unlimited | 5 reviews/hr rate limit | Flat seat price. No per-review overage. |
| GitHub Copilot Pro | $10 | PR review included | — | Cheapest by a wide margin. Bundled with Copilot. |
| Cursor BugBot Teams | $40 | Unlimited | — | Higher seat. Same vendor codes + reviews (separation-of-concerns risk). |
For a 20-engineer team writing at agentic volume, Greptile's effective cost lands roughly 3–5× CodeRabbit's flat rateand 5–8× Copilot's included Code Review. Sources, all live and verified 2026-04-30: greptile.com/pricing · coderabbit.ai/pricing · github.com/features/copilot/plans · cursor.com/pricing.
04 · HYPOCRISY
What They Said. What They Did.
Five claims Greptile (or its CEO) made publicly, paired with the receipt that contradicts each one. All quotes verbatim. Sources linked.
“Open source projects with MIT, Apache or GPL licenses are eligible for 100% off.”
@goetzrobin (Mar 9 2026, 11 likes): "I signed up as an OSS maintainer and got confirmation we are approved. I still got $180 bill which is more than the project gets through GitHub sponsorship. Support is unresponsive." / @daradoescode (Apr 16 2026, 8 likes): "I'm in the greptile oss program but I'm being charged $30."
Same OSS-program members billed despite explicit approval. Refunds happened only after public Twitter pressure (the @goetzrobin arc resolved Mar 12 2026 only after @muzzdotdev DM'd him).
“Without [the price hike], we wouldn't be able to continue using frontier models.”
CodeRabbit Pro ($24/seat flat), GitHub Copilot Pro ($10/seat with PR review included), and Cursor BugBot Teams ($40/seat unlimited) all run frontier models without per-review overages. The frontier-model defense is unfalsifiable when the entire competitor set proves it false.
“We're developers ourselves, so we hold our products to high standards. We want to build beautiful things that inspire us.”
No spending caps. No in-app cancel button. No proactive overage notifications until the bill arrives. Eleven separate accounts on X publicly report being unable to cancel without emailing support — exactly the dark pattern California BPC §17600 was written to prohibit.
“since greptile only charges on active usage, simply stopping usage would make it so you're not charged any more. there is no [cancel button needed]…”
California Automatic Renewal Law (BPC §17600 et seq.) requires online cancellation paths for any consumer subscription. "Just stop using it" is not an opt-out mechanism a regulator recognizes. The EU Consumer Rights Directive position is even stricter (per @fiddyresearch's note: "this kind of stuff cannot survive in the eu").
The marketing surface says developer-first, OSS-friendly, frontier-models-cost-money, we hold ourselves to high standards. The pricing surface shipped without spending caps, without an in-app cancel button, without grandfathering, while billing OSS-program members the program explicitly carves out. Refunds for the OSS-billing complaints happened only after public Twitter pressure. Refunds as a policy don't exist; refunds as PR triage do.
05 · LEAVING
There Is No Cancel Button.
The pricing change is one accountability question. The cancellation experience is a second. Multiple customers report no in-product cancel button — you have to email support. California Automatic Renewal Law (BPC §17600 et seq.) requires online cancellation paths for any subscription sold to California consumers. The pattern below is exactly what the law was written to prohibit.
“while i appreciate the quick cancellation, forcing an email to support to cancel is exactly the bad practice california business and pr…”
“Greptile doesn't let cancel subscriptions, either a bug or a scam. In any case, there is no customer support.”
“There's presently no area in billing to cancel your Greptile account btw. Just noticed that while looking for price change.”
“@greptile Why do I have to contact support to cancel my subscription? Did you forget this exists?”
“Literally had this happen to myself trying to cancel Greptile for our team.”
“if you ever decide to cancel your greptile subscription - you have to contact their support because apparently there is no…”
“Apparently to cancel a @greptile subscription you need to send them an email. this is the kind of stuff that just cannot survive in the eu.”
“i have spent 1 hour trying to cancel my @greptile subscription. still cant find a button or anything.”
“asking to contact support to cancel a subscription, while allowing to upgrade immediately is the reason I'll make sure I t[ell people…]”
“I've been sending emails to support for a week now with no answer.”
“I would like to reach out for some support for a billing I did not expect. How could I reach out? I did not find any mail/prop[er channel]…”
06 · THE LAUNCH POST
What The V4 Announcement Is Actually Doing.
Read the Greptile v4 launch post alongside the actual customer experience and four rhetorical moves stand out:
- The "90% won't see an increase" framing. On X (@dakshgup, 2026-03-05): "90% of users will not see their price increase at all."Without disclosing the user mix, this is structurally meaningless. The 10% who exceed are the customers who actually generate Greptile's usage and revenue. Punishing them is the opposite of "supporting power users."
- The "better coding agents" rationale. The blog post says: "Better coding agents have led to a drastic increase in the number of commits reviewed by Greptile."The acknowledgement is correct. The pricing response — capping at 50 and charging $1 per review thereafter — penalizes the exact use case the rationale invokes. The two cannot coexist without the cap functioning as a usage tax on the product's own stated growth driver.
- The unannounced switch. @CyprienHalle: "@greptile switched to usage-based pricing without notice lol → unsubscribe." Other customers report no email of the change. No grandfathering of existing flat-rate subscriptions was offered.
- The missing controls. @kris_remback: "why can't I set spending limits to avoid overages?"A usage-based pricing model that doesn't expose spending caps to customers is the shape of a vendor optimizing for upside, not a customer relationship.
Each move on its own is defensible. Stacked together they describe a pricing change designed to maximize revenue from the customers least able to forecast their bill — the same customers Greptile's product is built for.
07 · CUSTOMERS
Paying Customers Cancelling, In Public.
Seven of the strongest customer posts from X, March–April 2026. None solicited. All quoted verbatim with timestamp and link. The wave is two months old and surfaces weekly.
“this pricing change is a doozy. In my workflows, I put up smaller PRs in stacks, but that means more of them. $1/review past 50/mo will be blown past in less than 3 days. In the last 30 days, my agents have put up 571 PRs for just a subset of my repos. … going from $30/mo to more than $500/mo is not gonna work for me.”
“I think ill have to say goodbye to @greptile their new pricing sucks”
“I cancelled.”
“we cancelled greptile recently due to price increases, so we've cut down to Bugbot + Codex reviews now.”
“@greptile bill exceeding my claude code bill at this point.. 🙃”
“One day, $500 bill. Greptile used to be good.”
“You can easily hit 10 code reviews per PR with back and forth because greptile does not always catch all the issues in the first go.”
08 · VERDICT
A Pricing Model Designed For The Last Era.
The 50-review cap made sense when AI code review was a manual augment to human PR flow. It does not make sense when the vendor's own rationale for the change is "agents are writing more code," the vendor's own math says the cap covers ~42 PRs/dev, and Linear's company-wide median is already at 33. The two facts cannot coexist without the cap functioning as a usage-tax surcharge on the product's own stated growth driver.
The market answer, in real time, is cancellation. The product answer is bundling — flat seat pricing or per-org pricing that tracks engineering throughput, not commit volume. Until Greptile reverses course, the active market signal is switch to CodeRabbit Pro at $24/dev flat or GitHub Copilot Pro at $10/dev with PR review included and route the difference anywhere but Greptile.