Setup, compared for small teams
Skyvern is easy to use. “Easy to set up” is a longer checklist than that.
For a small business owner weighing AI workflow tools, the friendly authoring screen is the part everyone shows you. The part that decides whether you ship is the setup underneath it: who holds the LLM bill, what you have to install, and whether the tool can even reach the app your work lives in.
The short answer (verified 2026-06-17)
Skyvern is genuinely easy to author with, and its Cloud path is the fastest first run of any AI automation tool here. But “setup” for a small business still has two steps the friendly screen hides: you bring your own LLM provider API key (OpenAI, Anthropic, or Gemini, with its own billing), and if you self-host you add Docker, PostgreSQL, and Python 3.11. The bigger gate is reach: Skyvern only drives web browsers. If your work lives in a desktop app with no website, friendliness never gets the chance to matter.
Verified against Skyvern’s own quickstart and LLM-configuration docs at skyvern.com/docs/developers/getting-started/quickstart.
“User-friendly” and “easy to set up” are not the same question
When small business owners search for whether Skyvern is friendlier and easier to set up than other AI-driven automation tools, they are usually picturing one screen: the place you describe a task and it runs. On that screen Skyvern is strong. You write the task in plain language or record a procedure, and a vision model reads the live page instead of a brittle list of CSS selectors that breaks the moment a site is redesigned. That is real, and it is a genuine advantage over the selector-based scripting that older RPA tools lean on.
But authoring is the last step of setup, not the whole of it. Before you reach that friendly screen there is account creation, a model provider to fund, and (if you do not use Cloud) infrastructure to stand up. And before any of that matters at all, there is one yes-or-no question that most comparisons skip entirely: can the tool touch the software your business actually runs on. So look at all of it, not just the demo.
The real setup checklist, on Skyvern Cloud
The fastest honest version of Skyvern setup is the Cloud path. Here is what it actually takes for a small team to get to a first useful run. Step two is the one people do not expect.
Skyvern Cloud, start to first run
- 1
Sign up at app.skyvern.com
Create an account, copy your Skyvern API key from Settings.
- 2
Bring your own LLM key
Pick a provider (OpenAI, Anthropic, Gemini) and add billing. Skyvern drives the agent; the model bill is yours.
- 3
Describe or record the task
Plain language or a recorded procedure. A vision model reads the live page, no CSS selectors.
- 4
Run and watch the recording
First run on a public website often lands in minutes.
None of these steps are hard for someone comfortable with APIs. But step two, “bring your own LLM key,” means opening and funding a separate OpenAI, Anthropic, or Gemini account before the tool does anything for you, and that account, not Skyvern, is where your per-run model cost lands. For a developer this is five minutes. For an owner who expected one signup and one bill, it is a surprise worth knowing about up front.
If you self-host, here is what “quickstart” provisions
Skyvern is open source, and self-hosting keeps your data on your own infrastructure, which some small businesses want. The setup is a single command, but that command stands up more than a single app. Running skyvern quickstart provisions three services: a PostgreSQL database, the skyvern container that bundles the API server plus Playwright with Chromium, and the skyvern-ui frontend. The image is built on Python 3.11. It runs without a GPU.
You can verify every line of this in Skyvern’s own docs: the three-service Docker layout and Python 3.11 base in the Docker setup guide, and the mandatory LLM provider selection in the LLM configuration docs. For a small business with a developer on hand, this is a reasonable afternoon. For one without, the Cloud path exists precisely so you never touch any of it.
The question that outranks friendliness: can it reach your software?
Here is the decision that should come before any ease-of-use comparison. Skyvern is a browser agent. It drives websites and web apps, and it is good at that. It does not drive native Windows desktop software, green-screen terminals, SAP GUI, or any program without a web interface. A perfectly friendly tool that cannot open the application your work happens in is not an easy setup, it is a dead end you reach more pleasantly.
For a lot of small businesses the work is genuinely in the browser: web portals, web-based bookkeeping, SaaS dashboards. If that is you, stop reading comparisons and use Skyvern. But plenty of small businesses still run a desktop application with no API behind the counter or in the back office, an installed accounting program, a practice-management or point-of-sale system, a county or carrier portal that only exists as a Windows program. That is the fork.
Which path fits a small business
Where does the work live?
Be honest about the app you click into every day.
All in the browser
Web portals, SaaS, web bookkeeping. Skyvern Cloud is the faster, self-serve, lower-cost fit.
In a desktop app with no API
Installed Windows software, SAP GUI, green screen. No browser agent reaches it.
Accessibility-API automation
Mediar drives the desktop app you are already in, through the same APIs screen readers use.
How Mediar’s setup is shaped differently
Mediar exists for the second branch of that fork, so its setup is built around a different constraint. You download a Windows desktop app, or use the no-code recorder at app.mediar.ai/web, and it drives the applications you are already logged into through Windows accessibility APIs. There is no LLM provider key for you to bring and fund, no Docker, no PostgreSQL, and no selectors to maintain when a screen changes, because it reads what the app exposes to assistive technology rather than matching pixels.
The honest counterweight: Mediar is sales-led, not self-serve. You do not put a card in and start alone. You begin with a pilot, a turn-key program that gets a first workflow into production, and runtime is billed at $0.75 per minute. That shape fits a team with a recurring, high-volume legacy-desktop workload, not a solo owner automating a single web form. If you are the latter, the friendlier and cheaper answer really is a browser tool, and we would rather you hear that than buy the wrong thing.
One-line summary of the difference
Skyvern optimizes for a friendly, self-serve start on the web. Mediar optimizes for reaching desktop software that has no web at all. Pick by where your work lives, not by which demo felt smoother.
Not sure which side of the fork you are on?
Tell us the app you are trying to automate and we will say plainly whether a browser tool fits better or whether the desktop layer is your blocker.
Setup and ease-of-use questions
Is Skyvern actually easy to set up for a small business?
For the authoring experience, yes. You describe a task in plain language or record a procedure and a vision model reads the live page, so you do not maintain CSS selectors. Skyvern Cloud is the fastest first run of any AI automation tool in this comparison: sign up, copy an API key, go. The two parts that surprise small teams are that you still bring your own LLM provider key (OpenAI, Anthropic, or Gemini) with its own billing, and that self-hosting adds Docker, PostgreSQL, and Python 3.11. Neither is hard for a developer; both are real steps for a non-technical owner.
Do I need my own OpenAI or Anthropic API key to use Skyvern?
Yes. Skyvern's self-hosted quickstart explicitly asks you to select an LLM provider (OpenAI, Anthropic, Azure OpenAI, Gemini, or a local Ollama model) and paste a key, and the model usage is billed to that provider account, separate from Skyvern. This is normal for AI automation tools, but it means 'setup' includes opening and funding a second account before your first useful run.
What does self-hosting Skyvern actually install?
Running 'skyvern quickstart' provisions three services: a PostgreSQL database, the skyvern container that bundles the API server plus Playwright with Chromium, and the skyvern-ui React frontend. The Docker image is built on Python 3.11-slim-bookworm. It runs fine without a GPU. For a small business with no one comfortable in a terminal, the Cloud path avoids all of this.
Can Skyvern automate my desktop software, like QuickBooks Desktop or an ERP?
No. Skyvern is a browser agent. It drives websites and web apps. If your business runs in a native Windows desktop application, a green-screen terminal, SAP GUI, or any program with no web interface, no amount of Skyvern's friendliness reaches it. This is the gating question that matters more than ease of use: friendliness only counts if the tool can touch the place your work happens.
How is Mediar's setup different from Skyvern's?
Mediar takes a different path because it solves a different problem. You download a Windows desktop app or use the no-code recorder at app.mediar.ai/web, and it drives applications you are already logged into through Windows accessibility APIs, the same interfaces screen readers use. There is no LLM key for you to bring, no Docker, no PostgreSQL. The honest trade is that Mediar is sales-led: you start with a pilot, not a self-serve signup. It is built for teams whose work lives in legacy desktop systems, not for a solo owner automating a web form.
If my small business runs entirely in the browser, should I use Skyvern or Mediar?
Use Skyvern. If every system you touch is a web app and you want self-serve, Skyvern Cloud is the faster, cheaper, more appropriate fit, and we will tell you that plainly. Mediar earns its place when the work lives in a desktop application with no API, where a browser agent simply cannot go.
What does Skyvern cost compared to Mediar for a small team?
Skyvern Cloud is usage-priced and self-serve, plus the LLM provider bill you bring. Mediar is $0.75 per minute of runtime with a $10,000 turn-key program fee that converts to credits, and it is sold through a pilot rather than a card-on-file signup. For a small browser-only workload, Skyvern is the lower-friction, lower-cost option. For a recurring legacy-desktop workload that no browser tool can run at all, Mediar's number is the cost of work that otherwise stays manual.
Keep reading
Skyvern Cloud pricing in 2026
What Skyvern Cloud actually costs, and the LLM provider bill that sits on top of it.
Skyvern ease of use and 2FA, without the green checkmark
Friendliness is two questions. The second one, who holds your second factor, decides more than the first.
Legacy desktop apps with no API: the part browser agents miss
Why the desktop layer is where browser-based AI agents stop, and accessibility-API automation begins.
Comments (••)
Leave a comment to see what others are saying.Public and anonymous. No signup.