Kilo Code vs GitHub Copilot: the subscription is not the bill

Why this page exists this week: Kilo Code's native JetBrains agent was the number one launch on Product Hunt for the week of 31 August 2026, with 512 points and the tagline "fully native, open-source coding agent built for JetBrains". That framing invites an obvious conclusion, and the obvious conclusion is wrong.

First, the thing you were about to believe

GitHub Copilot already works in JetBrains IDEs. GitHub's own plans page lists "the JetBrains suite of IDEs" among supported environments, next to Visual Studio Code, Visual Studio, Vim, Neovim, Xcode, Eclipse, Zed, Raycast and more. A launch headline about native JetBrains support is a statement about Kilo's own product line, not a gap in Copilot's.

So do not choose on IDE coverage. Both cover JetBrains, and both cover VS Code. What actually separates these two is something no launch post leads with: who decides what your inference costs.

The two pricing models, side by side

Both companies publish their numbers, which is rarer than it should be. Read down the middle column and then the right one, because they are not the same kind of number.

Prices as printed on github.com/features/copilot/plans and kilo.ai/pricing, both read 7 September 2026. GitHub labels its figures USD on the page. Kilo prints dollar signs without a currency label, and we read the page from a US connection, so treat Kilo's as USD but confirm at checkout.

The $10 is partly a credit purchase

This is the detail that changes the argument, and almost nobody writing about Copilot mentions it. Copilot Pro costs $10 a month and includes $15 of monthly GitHub AI Credits. The pattern repeats: Pro+ is $39 and includes $70, Max is $100 and includes $200.

Which means the honest reading of Copilot Pro is not "a $10 subscription on top of your model costs". It is closer to a subscription that hands back more credit than it charges, with completions on top. If you had assumed Copilot was the expensive option because it has a price tag and Kilo does not, that assumption does not survive the plans page.

The mirror image is just as important. Kilo's $0 is real, and the zero markup on inference is a genuine, published, checkable commitment. But zero markup on a bill you generate yourself is not the same as a small bill. An agentic coding loop pointed at a frontier model can burn through more than $15 of tokens in a focused afternoon. Kilo's price is not low, it is uncapped, and uncapped runs in both directions.

What we cannot tell you, and will not pretend to

The question you actually want answered is which one costs less across a real month of your work. We are not answering it, because it cannot be answered from two pricing pages and we have not run the test.

To answer it honestly, somebody has to run both tools over the same real jobs for a fortnight and log what the meter says. That means measuring how much inference an agentic loop consumes against how much a completion-first assistant consumes, on the same tickets, in the same repositories. Until that is logged, any number we published would be invented. What we can say from the vendors' own pages is this: Copilot's monthly cost is knowable in advance and Kilo's is not. For some developers that predictability is worth $10 on its own. For others, being able to point the tool at a cheap model, or a local one, and pay nothing at all is worth more than the predictability.

Choose GitHub Copilot if…

• You want one line item and no surprises. $10 buys unlimited completions and $15 of credits, and your finance admin for the month is finished.
• You mostly want the assistant that finishes the line you are already typing, rather than an agent you brief and supervise.
• You bill clients for your time and would rather not spend any of it managing a model provider account, a key, and a usage dashboard.
• You want to start at zero risk: the free tier is 2,000 completions a month with no card required.

Choose Kilo Code if…

• You want to choose the model, not be given one. 500+ models across 60+ providers through the Gateway, or your own key, or something running locally on your own machine.
• Client contracts constrain where code can go. Being able to run a local model, or route to a specific provider you have already cleared, is a compliance answer rather than a preference. Settle that in the contract before the project, not after.
• You want the agent rather than the autocomplete, and you are comfortable supervising one.
• Open source matters to you, either on principle or because you want to read what the extension does before you install it in a client's repository.

One thing to watch, stated plainly

Multiple review sites report that Anaconda acquired Kilo Code in July 2026. We could not confirm that against a first-party statement from either company in the research for this page, so we are recording it as unverified rather than repeating it as fact. It matters only if it turns out to be true and the licensing or the free tier changes as a result. The extensions are open source today, which is the practical protection against that, but if you are about to standardise a team on this, spend five minutes confirming ownership yourself.

What we have not evaluated

This page compares exactly two tools, because Copilot is the coding-assistant pick in our freelance developers stack and Kilo Code is the credible new challenger to that slot this month. There are other agentic coding tools and other editor assistants. We have not evaluated them, so we are not going to rank them here.

We also have not run either tool across a real client project. Everything above comes from the two vendors' current published pricing and plan pages, read on 7 September 2026, which is why this is labelled a Research Edition and not a test result. Specifically unmeasured: output quality, agent reliability over long sessions, how the JetBrains extension behaves after a week of use, and the inference spend question that the whole cost comparison turns on. When the hands-on cycle runs, this page gets re-scored and the changelog records what changed.

See these tools inside the full stack: Freelance Developers stack → · related work: Solo Consultants → · Jasper vs Claude →

Common questions

Does GitHub Copilot work in JetBrains IDEs?

Yes. Kilo Code's JetBrains agent was the top launch on Product Hunt for the week of 31 August 2026, which makes it easy to assume JetBrains support is the thing Copilot lacks. It is not. GitHub's own plans page lists the JetBrains suite of IDEs among supported environments, alongside Visual Studio Code, Visual Studio, Vim, Neovim, Xcode, Eclipse, Zed and others. If you are choosing between these two, do not choose on IDE coverage. Both cover JetBrains.

Is Kilo Code actually free?

The editor extension is. Kilo's pricing page lists the open source VS Code, JetBrains and CLI extensions at Free, with the words "Usage is billed separately" printed directly underneath. That second sentence is the whole story. You are not paying for the tool, you are paying whichever model provider you point it at. Kilo charges no markup on that, and its own Gateway is listed at $0 a month plus usage at exact provider rates. So Kilo Code is free the way a kettle is free once you own the electricity.

What does the $10 for Copilot Pro actually buy?

More than it first looks. GitHub's plans page lists Copilot Pro at $10 USD per user per month with unlimited code completion and next edit suggestions, and it includes $15 of monthly GitHub AI Credits. So part of that $10 is a credit purchase at better than face value. Pro+ is $39 with $70 of credits and Max is $100 with $200 of credits, which is the same trade at larger sizes. Whether $15 of credits covers your working month is a usage question we have not measured.

Which one is cheaper for a freelance developer?

We will not answer that, because nobody can answer it from a pricing page and we have not run the test. Copilot Pro is a known $10 a month with a known $15 credit allowance inside it. Kilo Code is a known $0 for the editor and an unknown amount for inference, because that depends on which model you point it at and how hard you drive an agent. An agentic loop against a frontier model can spend real money in an afternoon. The honest statement is that Copilot's bill is predictable and Kilo's is uncapped in both directions, and where the crossover sits depends on a week of your actual work.

Do either of these pay ProvenStacks a commission?

Neither, today. GitHub Copilot has no affiliate programme at all, which is why it has been a $0 pick in our freelance developers stack since the page was built. Kilo does run a direct partner programme, published at kilo.ai/partners, paying a one-time 50 percent of the first month on its Kilo Pass credit tiers. We are not enrolled in it and we did not look at it until after the comparison above was written. If we ever enrol, that goes in the commission table and the changelog before it changes a word here.