Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Clear out junk files and repair common Windows errors3Scan for outdated or missing drivers - takes under a minuteiTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
Grunz is a desktop coding agent whose maker launched it as a way out of recurring AI subscriptions. The maker’s original post describes a one-time $100 purchase. Grunz’s current official documentation describes something different: prepaid, pay-per-use inference credits, billed at each model’s per-token price plus a flat 10% margin. The app also does not host a model. It runs file actions on your computer, but it sends your requests to an outside inference provider. This article separates the maker’s claims from the current documented setup so you can judge the product before you pay for it.
What the maker’s original post claims
The product was introduced in a first-person post on DEV Community, “I got tired of AI subscriptions, so I built a coding agent you buy once and self-host”. The author describes two frustrations that led to the project: a monthly bill that never stops, and a usage limit or refusal arriving at the moment the tool was most needed. That is the author’s own experience, not survey data about developers in general.
The post makes several claims that are useful to understand, though none of them were independently tested:
- Grunz is a native desktop agent for Mac, Windows, and Linux.
- You give it a goal, and it plans, writes and edits files across a repository, runs commands, watches the output, and attempts fixes.
- It includes cited web research and image generation.
- It has a tool loop of around 50 tools, a plan mode, and a permissions setting with ask-before-change, auto-accept edits, and full-auto options.
- The price is “$100 once,” with no subscription, and the user pays the model provider directly for compute.
The post sums up the economics in one sentence: “You pay your model provider directly for compute, and that’s the only ongoing cost — which means you control it.” Read that as the author’s description of the setup at the time of writing.
What Grunz is today, by platform
Grunz’s official documentation describes a desktop agent that edits files and projects, writes and runs code, generates images, and researches the web. The homepage at grunzai.com lists the download surfaces. The minimum requirements are below. They can change with new builds, so check the docs before installing.
| Platform | Requirement stated in current official material | Source |
|---|---|---|
| macOS | macOS 12 or later on Apple Silicon. Intel Macs are not listed in the docs reviewed for this article. | Grunz Docs |
| Windows | Windows 10 or 11, 64-bit. | Grunz Docs |
| Linux | Modern Linux with AppImage support. The homepage also names Linux x86_64. | Grunz Docs; Grunz homepage |
| All platforms | An internet connection. No local model hardware is listed as a requirement. | Grunz Docs |
The absence of a GPU or local-model requirement matters for the “self-host” question covered below. Grunz’s requirements describe a machine that can run the app and reach the internet, not a machine that runs an AI model.
Pricing: the original claim versus current documentation
This is the most important point for anyone deciding whether to buy. The original post and the current official documentation describe different pricing arrangements, and the post’s “$100 once” figure should not be presented as the current price.
| Item | Original DEV Community post | Current Grunz documentation and Terms |
|---|---|---|
| Upfront charge | “$100 once” (the author’s stated price) | No license fee or monthly subscription. Users load prepaid credits starting at $1. |
| Model cost | User pays the model provider directly for compute. | Charged the model’s per-token price plus a flat 10% margin. The Terms say provider costs are passed through with that margin. |
| Ongoing cost | Described as “the only ongoing cost.” | Usage-based. Cost rises with use. |
| Source and date | Author’s post, at publication | Grunz Docs; Grunz Terms of Service, version 2026-09-26 |
Grunz’s documentation states the model plainly: “You are charged the model’s exact per-token price plus a flat 10% margin. That is the whole pricing model.” That sentence comes from Grunz itself and has not been independently verified.
No source reviewed for this article confirms that a perpetual $100 license is currently offered. Treat the original post’s price as a historical claim unless the vendor confirms it. Do not assume either arrangement without checking the current purchase flow.
Buying details that the pricing model implies
- Credits are tied to where you buy them. Grunz says credits live in the browser or installation and do not sync across devices.
- Credits are prepaid and generally non-refundable. The Terms describe them this way. Read the refund language before you top up.
- The 10% margin is a flat rate on top of provider token cost. The Terms say so, but the per-token prices themselves depend on the model you choose and are set by the provider, not by Grunz.
- Usage is metered. Grunz says it records token counts and costs to meter requests, so your spend tracks how much the agent works.
What “self-host” means here
The title’s phrase “self-host” is the point most likely to mislead. The original post itself says the desktop app does not host a model. It connects to a model endpoint. The current Grunz Terms are more specific: Grunz is a pass-through interface, and it says requests go to an independent inference provider, currently Featherless.ai. The Terms also state, “Grunz does not create, train, fine-tune, modify, or host any AI model.”
Here is the data path as the current documentation describes it:
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstall- The agent runs on your computer and performs file actions such as reading, editing, and writing project files there.
- When a request needs a model, file content included in that request is sent through Grunz’s routing layer to the inference provider.
- The provider returns the model’s output, and Grunz applies the file actions on your machine.
- Grunz says it does not store prompts, files, or generated content. It keeps token counts and costs for metering.
In practical terms, the app is local, but the thinking is not. If your code is sensitive, the important question is whether the provider’s terms and retention policies are acceptable to you, not whether the app runs on your laptop. The provider’s pricing and plan details are published on Featherless Pricing and Featherless Plans. Those pages are volatile, so check them directly before relying on any figure.
Privacy statements to weigh
- Grunz’s statement: it does not store prompts, files, or generated content.
- Limited records the Terms allow: records may be kept for metering, payment, abuse prevention, and recording acceptance of the Terms.
- What is not established: the sources reviewed do not include an independent audit of these statements, and they do not describe the provider’s own retention practices in full.
Agent safety and your control over actions
A desktop agent that edits files and runs commands can cause real damage, and Grunz’s Terms say so directly. The Terms describe actions that can read and edit files, write or run commands, browse, and generate images. They warn that generated commands or actions may be wrong, destructive, or cause data loss or vulnerabilities. This is the vendor’s contractual warning, not a report that a particular incident has occurred.
The original post describes a permissions setting and a plan mode. Those controls were described by the author and were not tested for this article. Treat them as claims until you confirm them in your installed version.
Based on the Terms, a careful setup looks like this:
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
- Work in a version-controlled project so every change can be reviewed and reverted with your normal tools.
- Keep a tested backup of anything the agent can touch, not just an untested copy.
- Run the agent in an environment suited to the risk, such as a sandbox, virtual machine, or container, especially for unfamiliar tasks.
- Start with the most restrictive approval mode and review each proposed file change and command before approving it.
- Do not grant full-auto operation on a repository you cannot easily rebuild.
How to compare any buy-once or self-hosted coding agent
Grunz is one option among several AI coding tools. This article does not rank it against competitors, because no independent comparison was reviewed. Instead, use these six axes to check any product that claims to be a one-time purchase or self-hosted:
- Billing model: is it a fixed subscription, metered inference, or a desktop license? Are there charges beyond the stated price?
- Where inference runs: is the model local or remote? Which provider receives your data, and what do its terms say about retention?
- Platform support: which operating systems and CPU architectures are listed, and what install format is used?
- Agent controls: what permission modes exist, which commands can run without approval, and how are changes reviewed or reverted?
- Model and limits: which models are available, what context size applies, and what request or rate limits are published?
- Cost measurement: how are credits or usage counted, and do they carry across devices or expire?
Applied to Grunz, the current evidence supports these answers: remote inference through Featherless.ai as the named provider, file actions on your device, and provider token cost plus a flat 10%. The billing model, the one-time price, and the agent controls are the points most in need of confirmation before you spend money.
What to confirm before you buy
- Ask Grunz whether the “$100 once” offer described in the original post is available now, and what it includes. Current documentation does not describe it.
- Read the current Terms of Service, version 2026-09-26, for refund, credit, and data-handling language.
- Confirm your operating system and CPU architecture against the requirements table above.
- Test the permission modes on a throwaway project before trusting them with real work.
The maker’s motivation is clear and understandable. The product as currently documented is a metered, pass-through agent with local file actions. Whether that suits you depends on the billing details and the data path, not on the “buy once” framing.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
