Economics

Flox has three published plans: a free Personal plan, a Team plan at $20 per seat per month, and Enterprise with custom pricing. Flox’s own materials describe the cost of environment drift in lost hours but publish no percentage, so any drift-cost number in a deal is the customer’s own estimate. This page lists every published price and limit, derives seat costs only from the published price, and explains how to use the seat-cost versus drift-cost calculator without inventing defaults.

This page is drawn from https://flox.dev/pricing/, https://flox.dev/docs/concepts/organizations, https://flox.dev/docs/concepts/floxhub, https://flox.dev/blog/flox-catalog-login-requirement-what-you-need-to-know/, https://flox.dev/case-studies/how-fellow-ai-ships-faster-with-flox/, https://flox.dev/case-studies/how-resolve-ai-eliminated-works-on-my-machine/, https://flox.dev/blog/works-on-my-machine-a-platform-problem/, https://flox.dev/control-your-sdlc/, https://flox.dev/kubernetes/, https://flox.dev/blog/deploying-hardened-flox-nvidia-cuda-stacks-in-minutes-not-hours/ and https://flox.dev/cuda/, all checked 2026-11-04. Lines marked derived are arithmetic on those numbers, not quotes from Flox. Prices are USD as shown on the pricing page.

What Flox charges

The pricing page lists three plans (https://flox.dev/pricing/). The first is labeled Personal in the page text and priced “Free”; this guide calls it the Free plan (https://flox.dev/pricing/).

Three things to say out loud:

• The Free plan is a real product, not a trial. The pricing page describes it as for “developers who want to build and share portable, reproducible dev environments using a large catalog of open source software” (https://flox.dev/pricing/).
• The limits that bite first are the private ones: 3 private FloxHub envs and 5 private packages on the Free plan, versus unlimited team FloxHub envs and unlimited publishing of packages on Team (https://flox.dev/pricing/). Public FloxHub envs are unlimited on the Free plan (https://flox.dev/pricing/).
• Organizations are a Team feature. The docs mark organizations as a “Flox for Teams” paid feature and describe an organization as providing a private catalog, scoped access control and a foundation for collaboration (https://flox.dev/docs/concepts/organizations). Automated environment upgrades on FloxHub are also marked as included with Flox for Teams (https://flox.dev/docs/concepts/floxhub).

What the pricing page does not say: whether there is annual billing or a discount, a minimum seat count for Team, what “platform support” means in response-time terms, or the price of the consulting. Mark each [verify] and ask Flox rather than guessing (https://flox.dev/pricing/).

The login change is not a price change

Since 2026-10-20, installing, upgrading or resolving packages from the Flox Catalog requires a free FloxHub account (https://flox.dev/blog/flox-catalog-login-requirement-what-you-need-to-know/). Flox states: “This isn’t a paywall.” and says FloxHub accounts are free and the free tier keeps all its existing capabilities (https://flox.dev/blog/flox-catalog-login-requirement-what-you-need-to-know/). The stated reason is that agents generate more Catalog traffic and authentication helps curb abusive automation (https://flox.dev/blog/flox-catalog-login-requirement-what-you-need-to-know/).

Worked examples: seat cost (derived)

Inputs: Team plan $20/seat/month (https://flox.dev/pricing/). Every figure below is derived by multiplying that one published price by a seat count and by 12 months. Whether Team can be bought monthly without a minimum is [verify] (https://flox.dev/pricing/).

Formula: seats x $20 = monthly cost, using the price at https://flox.dev/pricing/. At 200 seats, a customer is more likely to be talking about Enterprise features (custom base catalog, software policy engine), and Enterprise pricing is “Custom pricing” with no published number (https://flox.dev/pricing/). Do not extrapolate the $20 figure into an Enterprise quote.

Who needs a seat is a scoping question. The pricing page describes Team as for teams who need one way to build, publish and manage software “across local dev, CI, and production” (https://flox.dev/pricing/). Whether a CI runner or a service account consumes a seat is not stated on the pricing page [verify] (https://flox.dev/pricing/). CI does not need a login if the lockfile is committed, because an environment with a valid lockfile activates without one (https://flox.dev/blog/flox-catalog-login-requirement-what-you-need-to-know/).

What the sources say about time lost to drift

Flox’s own claims are qualitative. None of the pages below publishes a percentage or a dollar figure for time saved, and this guide does not add one.

• Flox’s platform-team page says “The bill arrives as lost hours, and it lands on the platform team.” (https://flox.dev/control-your-sdlc/). The same page walks through a scenario that ends “Half a day gone, to a change that was one line.” (https://flox.dev/control-your-sdlc/). That is a narrated scenario, not a measurement [verify] (https://flox.dev/control-your-sdlc/).
• The same page says onboarding goes “days to minutes” (https://flox.dev/control-your-sdlc/). The works-on-my-machine post says “New engineers spend days assembling local environments from READMEs, Slack threads, teammate advice, package manager errors, and stale assumptions.” (https://flox.dev/blog/works-on-my-machine-a-platform-problem/).
• Fellow.ai (case study, 11 May 2026): the CTO says “people end up wasting time on ‘works-on-my-machine’ issues instead of working on product features” without consistent local dev (https://flox.dev/case-studies/how-fellow-ai-ships-faster-with-flox/). After rollout: “Once all the dev machines could run the same Flox environments, this rarely happened” (https://flox.dev/case-studies/how-fellow-ai-ships-faster-with-flox/). The case study says Flox “did essentially eliminate drift in local development” while not completely solving the problem (https://flox.dev/case-studies/how-fellow-ai-ships-faster-with-flox/).
• Resolve AI (case study, 1 June 2026): “Work stops while you figure out whether you’re looking at a regression in the code or just a divergence in somebody’s local environment.” (https://flox.dev/case-studies/how-resolve-ai-eliminated-works-on-my-machine/). After adoption, works-on-my-machine issues turned into “are you using Flox?” (https://flox.dev/case-studies/how-resolve-ai-eliminated-works-on-my-machine/). The case study also admits some engineers still avoided automatic activation with direnv when rebases triggered extra setup time (https://flox.dev/case-studies/how-resolve-ai-eliminated-works-on-my-machine/).
• Flox’s platform-engineer page narrates a week-by-week adoption story that includes “Pipeline time drops to 45 seconds” from 8-minute Docker builds and a patch “in 2 hours instead of 2 weeks” (https://flox.dev/platform-engineers/). These are illustrative scenarios on a marketing page, not customer data [verify] (https://flox.dev/platform-engineers/).

Say: “Flox’s published proof is two customer stories and a set of scenarios, all qualitative. The number that matters is yours: how many hours a month your team loses to setup and mismatched environments.” Ask for that number; do not supply one.

Image storage and pull costs (imageless Kubernetes)

Flox’s Kubernetes page says imageless Kubernetes lets pods “Pull declarative Flox environments at runtime” so each node downloads only the required hash-pinned packages, and describes this as skipping registry round trips (https://flox.dev/kubernetes/). It shows three headline figures: 800 to 0 MB under “Base image elimination”, 50% under “Faster provisioning”, and 5 to 2 under “Pipeline stages eliminated” (https://flox.dev/kubernetes/). The page gives no test setup behind them, so treat them as vendor claims [verify] (https://flox.dev/kubernetes/).

What the docs add:

• A pod names the Flox environment in an annotation and still needs a container image field. Flox uses a placeholder, flox/empty:1.0.0, which the docs call “only a few bytes in size” (https://flox.dev/docs/imageless-kubernetes/intro).
• That placeholder image is pulled from Docker Hub, and the environment itself is pulled from FloxHub (https://flox.dev/docs/concepts/floxhub).
• Imageless Kubernetes needs a containerd shim on each node and a minimum containerd version of 1.7 (https://flox.dev/docs/imageless-kubernetes/install/self-managed). It is available for Amazon EKS and self-managed Kubernetes (https://flox.dev/docs/imageless-kubernetes/intro).
• Updates do not rebuild an image: install the package with flox install -r myorg/myenv somepackage, run flox push -r myorg/myenv, and the next pod deployment pulls the new generation (https://flox.dev/docs/imageless-kubernetes/intro).

No Flox source gives a dollar figure for registry storage or pull traffic, because those are the customer’s registry vendor’s prices. For the calculator, the reader enters their own registry bill; Flox’s sources only say what is no longer pushed or pulled. Whether imageless Kubernetes is covered by the Free or Team plan, or needs Enterprise, is not stated on the pricing page [verify] (https://flox.dev/pricing/).

GPU and CUDA setup time (as sourced)

• Flox says that downloading a prebuilt, CUDA-accelerated PyTorch wheel from the Flox Catalog “takes 1m and 11s” while building it from source takes “55x longer (more than 65 minutes)” on a 32-core AMD ThreadRipper machine (https://flox.dev/blog/deploying-hardened-flox-nvidia-cuda-stacks-in-minutes-not-hours/).
• On the same machine, compiling a complete PyTorch inferencing stack “takes more than two hours” (https://flox.dev/blog/deploying-hardened-flox-nvidia-cuda-stacks-in-minutes-not-hours/).
• The earlier launch post says building packages such as tensorrt, opencv or ffmpeg with CUDA from source “could take hours”, and a full CUDA-accelerated ML stack “could take days” (https://flox.dev/blog/the-flox-catalog-now-contains-nvidia-cuda/).
• The CUDA page FAQ says the service is “completely free with no subscription or authentication required” (https://flox.dev/cuda/). The login post says installing from the Catalog needs a FloxHub account since 2026-10-20 (https://flox.dev/blog/flox-catalog-login-requirement-what-you-need-to-know/). These two pages do not agree on authentication [verify]: treat the login post and the docs as current and ask Flox whether the CUDA page is stale.
• GPU drivers are not included; they “remain the responsibility of the host operating system” (https://flox.dev/cuda/).

These timings are build-time measurements on one named machine, not a saving per developer. They matter for ML teams that rebuild CUDA stacks often; they do not translate to the seat calculator unless the customer actually builds from source today.

How to use the seat-cost versus drift-cost calculator

The calculator compares one published number with numbers only the customer can supply. Every input except the seat price is the reader’s own assumption, and no default is a fact.

Derived formulas, using the published seat price from https://flox.dev/pricing/:

• Seat cost per month = N x 20.
• Drift cost per month = N x hours lost per engineer x loaded hourly cost.
• Avoided cost per month = drift cost x share removed.
• Break-even hours per engineer per month = 20 / (loaded hourly cost x share removed). Example of the arithmetic only: if a customer enters a loaded hourly cost of 100 and a share of 0.5, break-even is 20 / 50 = 0.4 hours per engineer per month. Those two inputs are placeholders to show the division, not data.

Rules for using it in a conversation:

• Ask the customer to fill the blanks in front of you. Never pre-fill hours lost from a case study; the case studies publish none (https://flox.dev/case-studies/how-resolve-ai-eliminated-works-on-my-machine/).
• Show the break-even line first. It answers how little Flox would have to save to cover its own cost, with only one published input.
• Keep onboarding separate from steady-state drift, since Flox’s own page separates them: onboarding is the “episodic cousin” of the recurring drift tax (https://flox.dev/control-your-sdlc/).
• Do not count consulting or Enterprise features unless the customer is buying them; their prices are not published (https://flox.dev/pricing/).

Every price and limit figure, with its URL