Cursor and co. for SMEs: tools for people who code

Someone in your business may already have built a small tool with Cursor: a form, a report, a reconciliation between two lists. That is impressive, and it shows that software is much cheaper to make than it used to be. The open question is what happens next.

This article belongs to our series of comparisons: the same six questions for every tool. We read the providers’ information on 14 October 2026; the sources are listed at the end.

What you use today

Cursor is an editor with an AI agent (An AI agent carries out several steps of a defined task. A person decides before it sends anything outside your firm or spends money.). On 14 October 2026 its pricing page and its documentation list different plans: for teams it is USD 30.– or 40.– per person per month, depending on the page, plus usage for other providers’ models at their list prices. Cursor says daily agent users should expect USD 60.– to 100.– of usage a month. GitHub (A Git repository is storage for files that records every change with its date and author; GitHub is a well-known service for it.) Copilot (An AI assistant is a chat program such as ChatGPT, Claude, Gemini or Perplexity: you write in plain language, and a language model answers.) Business costs USD 19.– per person with a credit for AI requests. Windsurf is now called Devin Desktop. All prices in USD excluding tax.

Where Cursor and co. are strong

For people who code, these tools are a big leap. Cursor’s agent changes files, runs commands (A command starts a defined workflow. In Pro Setup, we set up three to five commands for your work.), searches the code and the web and operates a browser. Cloud (The cloud means computers in a provider’s data centres on which programs run and data is kept, instead of on your own device.) agents work in parallel in sealed-off virtual machines, Bugbot reviews pull requests, and rules in the project tell the agent how your team works. GitHub Copilot brings agent mode, code review and a cloud agent straight into GitHub.

Both speak MCP (MCP (Model Context Protocol) is an open standard for connecting an AI harness to other software and data sources.), the open protocol through which AI uses tools: Cursor through a configuration file, with OAuth sign-in for remote servers (A server is a computer that runs all the time and provides services to other devices, such as email, files or a website.), GitHub Copilot in the editor, the command line (The command line, also called the terminal, is a window in which you give the computer instructions by typing text instead of clicking.) and the cloud agent. In privacy mode, Cursor does not use your code for training, and no data is retained at the model providers (An AI provider runs language models and sells access to them, for example OpenAI (ChatGPT), Anthropic (Claude) or Google (Gemini).). Cursor has SOC 2, ISO 27001 and ISO 42001; GitHub does not train on data from Business accounts.

If you have one or two people who code, give them one of these tools. It pays for itself in days.

Where Cursor and co. stop

• They are for people who code. Cursor calls itself a coding agent for ambitious software. Anyone in the team who does not code gets little from it directly.
• After the code comes operations. A tool needs hosting (Hosting means that a provider runs the computers on which a website, a program or data runs and can be reached.), sign-in, a database, backups (A backup is a copy of your files in a second place, from which you restore data that was lost or damaged.) and updates. For publishing, Cursor points to partners such as Vercel (Vercel is a US provider that runs websites and their server functions.) or Netlify; its own code hosting is in early beta. Whoever built the tool usually ends up running it.
• Costs fluctuate. Besides the seat you pay for usage. Per-person spending limits come only with Cursor Enterprise, as does control over which MCP servers are allowed; on GitHub Copilot, MCP is off by default for organisations. With five people working with the agent every day, that quickly adds up to several hundred dollars a month.
• The data goes through the US. With Cursor, requests always pass through its servers, even with your own key; data is kept in the US, and there is no Swiss region (For cloud providers, a region is the area where the data centres that process your data are located, such as the EU or Switzerland.).

Cursor connected to our platform

The first way keeps your developers in their editor and gives them a foundation. Cursor, GitHub Copilot or Claude Code (Claude Code is an AI harness made by Anthropic: it runs on your computer, reads files and carries out steps you describe in plain language.) connect to the Jules Solutions Platform through its MCP gateway. Instead of rebuilding sign-in, document processing, messaging, schedules and a database for every small tool, the code uses what the platform already has. That leaves more time for what only your people can build.

The tools run on the platform in sealed-off areas, on our own infrastructure in Switzerland. You choose what to publish: a public link shows only what you chose, never the private vault (The vault is a folder of your working files on your computer; we create it from the vault template.). The code lives in your Git repository and belongs to you; agents propose changes, and a person accepts them.

For five people: Operator System (Operator System is our package for teams at CHF 3'500.– to 5'000.–: the shared base with workflows for the team and connected systems.) as the shared base, CHF 3'500.– to 5'000.– once, with six months of Business seat (The Business seat is one person’s access to our platform (open alpha); Pro Setup includes three months, and with Operator System it is six per person.) per person; then CHF 199.– per person per month, or CHF 169.– billed yearly. You keep paying for Cursor or Copilot.

The whole system

In the whole system, the people who do not code benefit too. Each person has a vault with working files and instructions, organised by PARA (PARA organises storage into four parts: projects, areas, resources and archives.) and readable in Obsidian (Obsidian is a note-taking program that shows a folder of text files on your computer as linked notes.), and starts workflows (A workflow specifies what information goes in, what happens to it and who checks the result.) with three to five commands, without writing a prompt (A prompt is the input to a language model: the question, instruction or template the AI responds to.). What the developers build reaches everyone as a tool, with sign-in, approvals and a log (A log records what a program or an AI agent did and when, so that afterwards you can trace who triggered what and what happened.). Whoever needs something of their own gets it built by us through Build It (Build It is one of our five ways of working: build what does not exist yet, around the way your work actually happens.).

An AI harness (An AI harness is the workspace on your computer in which AI reads files and carries out agreed steps.) works in the vault; we recommend Claude Code. Workflows run on your device, on a computer in the office, on our platform or with a cloud provider, as the quote sets out. We choose the language model (A language model is the actual AI behind an AI assistant: it has learned from a great deal of text and writes an answer to an input.) with you: in the cloud with providers in the US or the EU, or as an open-weight model (An open-weight model is a language model whose weights are freely available, so it can run on your own hardware.) on your own hardware.

The platform is in open alpha (Open alpha means the platform is at an early stage: we use it ourselves, and it changes all the time.); we run our own company on it. For five people: Operator System plus one Pro Setup (Pro Setup is our recommended package at CHF 1'490.–: vault, workflows with AI agents, first integrations, an introduction session and three months of Business seat.) per person, CHF 10'950.– to 12'450.– once, with 60 days of follow-up support (Follow-up support means that questions and adjustments by email and phone are included for 30 days (Pro Setup) or 60 days (Operator System) after handover.), then the Business seats and model usage (Model usage means accessing an AI provider’s language model. The cost is not included in the setup price.).

Our verdict

Do you have people who code? Then give them Cursor or GitHub Copilot, today. There is hardly an expense that pays for itself faster.

But every tool also brings a bit of operations: hosting, sign-in, backups, updates. The first way takes that foundation off your developers, so they build instead of run. The whole system pays off when the rest of the team should benefit from what gets built, and when you would rather have a partner for your own software than another building site.

What to settle before building is covered in ‘Five questions before you commission AI automation’.