Verdict up front: Kilo Code is genuinely free to start. It’s an open-source AI coding assistant with a $0 individual plan, and its Auto Free tier (kilo-auto/free) routes your requests to free models at no cost — limited to 200 requests per hour per IP. Paid options are optional: Kilo Pass (from $19/mo), pay-as-you-go credits, and cloud compute features. Be aware that autocomplete and background tasks use credits by default unless you switch them to free models. If you’re a complete beginner to AI coding tools, here’s your takeaway in one sentence: install Kilo Code, sign in, pick Auto Free, and start building — stay on free routes until rate limits or capability needs actually bite, and only then consider paying. Teams should look at the Teams plan at $15/user/month. This guide walks you through every step in plain language, no prior AI-tool experience required.
What is Kilo Code and can you really use it for free?
Yes — Kilo Code is an open-source AI coding assistant that works in VS Code, JetBrains, a terminal CLI, Slack, and iOS/Android apps, and the individual platform plan costs $0 (Kilo pricing). Its Auto Free tier routes your coding requests to the best available free models at no charge, so you can genuinely run an AI coding agent without paying anything, though the free models on offer change over time.
Let’s unpack that, because “AI coding assistant” covers a lot. Kilo Code is agentic, meaning it doesn’t just suggest the next line of code — it can hold a conversation about your project, generate and refactor code, debug issues, and run tasks like automated code reviews. It works wherever you do: as an extension in VS Code and JetBrains, inside Cursor, Windsurf, and other VS Code forks, from the CLI in your terminal, in Slack, and on your phone via the iOS/Android apps. The docs list a browser version as not yet available, so don’t count on a web app yet. Your sessions sync across all of these, so you can start a task on your phone and finish it in your IDE.
Because it’s open source, you can inspect the code and contribute to it. That openness matters for beginners in a practical way: you’re not locked into one company’s walled garden, and Kilo works with any AI model rather than only its own.
The free part works like this: Kilo separates platform access (the tool itself) from AI inference (the actual model calls that do the work). The platform is free for individuals. For inference, you can use Auto Free — a Kilo-provided tier that automatically routes your requests to the best available free models through Kilo Gateway where available, no configuration needed. Auto Free requests are billed at $0 on Kilo’s side, with a limit of 200 requests per hour per IP (rate limits and costs docs). One important caveat we’ll return to later: free routes may pass your code through providers that log prompts, so keep confidential work off Auto Free.
The decision rule for this guide: stay on Auto Free (or bring your own keys) until you hit rate limits or need a more capable model. Upgrade only when that actually happens — not before.
How This Guide Was Built
This guide was assembled from Kilo’s official documentation and pricing pages: the Kilo pricing page, the getting-started docs, and the specific pages on installing, authentication, using Kilo for free, and rate limits and costs. Every factual claim about plans, prices, limits, and features was checked against those pages, and nothing here comes from rumor or third-party speculation. To be fully transparent: this guide was not tested hands-on — we read the official sources carefully rather than running the tool ourselves, so treat the walkthrough as documentation-based rather than a personal playthrough. Last verified: November 2026. If Kilo changes its plans or free-model lineup after that date, the official pricing page is always the source of truth.
How do you install Kilo Code and run your first task?
Install the extension or CLI, then sign in — email, ChatGPT, Google, GitHub, and several other options all work, and signing in automatically configures everything you need to start (authentication docs). From there, follow the docs’ Quick Start to run your first task. The numbered steps below cover the whole path, including the free-model setup.
Getting started, step by step
-
Install Kilo Code. Pick your environment: the extension for VS Code or JetBrains, or the CLI for your terminal. You’ll find install instructions on the installing page of the docs. If you already use Cursor or Windsurf, the same extension runs inside those VS Code forks too.
-
Sign in or create a free account. When you first open Kilo Code, you’ll be prompted to sign in. You can use your email address, ChatGPT, Google, Apple, GitHub, GitLab, Discord, LinkedIn, or Anaconda — or your organization’s single sign-on. Signing in automatically configures everything you need to get started.
-
Run your first task. Follow the Quick Start: give the agent something simple to do, like explaining a file in your project or writing a small function. The point here is just to see the agent work end to end.
-
Select Auto Free in the model picker. Click the current model below the chat window and choose Auto Free (
kilo-auto/free). This tells Kilo to route your requests to free models automatically, so your first real usage costs nothing (using Kilo for free). -
Set the small model to a free model too. Kilo uses a small “background” model for things like session titles and context summarization, and by default that consumes credits. In the VS Code extension, go to Settings → Models and change the small model to any free model. In the CLI, set the
small_modelparameter in~/.config/kilo/config.json. -
Optionally bring your own key (BYOK). If you have an API key from OpenRouter, Anthropic, OpenAI, or another provider, open Settings → Providers, select the provider, and paste your key. The docs include step-by-step key-creation instructions for OpenRouter, Anthropic, and OpenAI. BYOK requests are billed at $0 on Kilo’s side — you pay the provider directly.
-
Configure by asking the agent. This is the easiest trick in the whole setup: just tell the agent what you want, like “add my Ollama endpoint” or “disable OpenAI.” The agent has a built-in skill for reading and updating your
kilo.jsoncconfiguration file (settings docs).
Pro tips
- Hit a rate limit? Switch free models. Free models are limited to 200 requests per hour per IP, and some upstream providers add their own limits. If you get throttled, type
freein the model picker to see every model labeled “(free)” and switch to another one. In the CLI, use the/modelscommand and filter the same way. - Chat your configuration instead of hunting through settings. The agent can edit
kilo.jsoncfor you, so plain-English requests like “add this MCP server” or “use my Ollama endpoint” replace most manual configuration. - Keep your context lean to stretch any model further. Write concise prompts, use
@fileand@foldermentions only when directly relevant, and break big tasks into smaller focused ones. Clear, focused prompts cost fewer tokens, and smaller sub-tasks are easier for the model to answer precisely (rate limits and costs docs).
How much does Kilo Code cost in 2026: plans, credits, and fees?
Kilo splits pricing into three parts: the platform plan (free for individuals, Teams at $15/user/month, Enterprise custom), AI inference (free models, BYOK, pay-as-you-go Gateway, or Kilo Pass from $19/mo with up to 50% bonus credits), and cloud compute billed per second (Kilo pricing). There is also a 5% processing fee when you purchase credits.
Part one: the platform plan. The individual plan is free and open source, covering the VS Code, JetBrains, and CLI extensions, the agentic assistant, code completion, commercial use of a large package library, and compatibility with any AI model. Teams costs $15/user/month and adds usage analytics, shared custom agents, centralized billing, shared BYOK, data privacy controls, and priority support. Enterprise is custom-priced and adds governance features like SSO, audit logs, model and provider access controls, and SLA commitments. Notably, the 14-day team trial is an Enterprise trial — so your team can evaluate Enterprise features before choosing between Teams and Enterprise.
Part two: AI inference. This is the part most beginners conflate with the platform, so read carefully:
- Auto Free (
kilo-auto/free): routes to the best available free models automatically, no configuration, billed at $0 on Kilo’s side. Limited to 200 requests per hour per IP. - BYOK (bring your own key): connect Anthropic, OpenAI, Google, Azure, AWS Bedrock, OpenRouter, and other providers. Billed at $0 on Kilo’s side — you pay the provider directly, at their rates.
- Local models: run models on your own machine with Ollama or LM Studio. Also $0 on Kilo’s side.
- Kilo Gateway: pay-as-you-go hosted inference with no subscription, giving access to 500+ models across 60+ providers at exact provider rates with no markup. You top up credits only when you need them.
- Kilo Pass: a monthly credit subscription starting from $19/mo, with up to 50% bonus credits on top of your subscribed credits. It’s a monthly AI credit subscription for regular usage.
Part three: cloud compute. Kilo’s cloud features — like Code Review and cloud agents — are billed separately, per second, with no rounding up or minimum usage charge. For example, Code Review runs at $0.33/hour, and cloud agents range from $0.60 to $1.20/hour depending on size (cloud pricing). Model inference is charged separately from this.
The fee to remember: purchasing Kilo credits carries a 5% payment-processing fee on top of the credit amount. It’s small, but if you’re budgeting precisely, factor it in.
One more thing worth knowing: Kilo’s Auto Model system routes each request to an appropriate model strategy based on the tier you pick — paid tiers trade cost against capability, and Auto Free rotates through whatever free models are currently available. The underlying models behind each tier are updated server-side as providers change, and you can see current assignments and live pricing at kilo.ai/models.
Kilo Code vs alternatives: which should you pick?
Kilo Code is open source and has a true $0 individual plan, so it is a sensible first pick for beginners who want to learn without spending (Kilo pricing). Cursor, GitHub Copilot, and Claude Code suit different workflows. The table below stays qualitative, and our comparison hub has deeper head-to-head breakdowns.
| Feature | Kilo Code | Cursor | GitHub Copilot | Claude Code |
|---|---|---|---|---|
| Open source | Yes — inspect, contribute, or fork it | Check the vendor’s site | Check the vendor’s site | Check the vendor’s site |
| Starting price | $0 platform plan; inference can be free | Check the vendor’s pricing page | Check the vendor’s pricing page | Check the vendor’s pricing page |
| Free tier for agentic coding | Yes — Auto Free routes to free models, up to 200 requests/hour per IP | Check the vendor’s pricing page | Check the vendor’s pricing page | Check the vendor’s pricing page |
| BYOK / local models | Yes — your own provider keys, plus local models via Ollama or LM Studio | Check the vendor’s docs | Check the vendor’s docs | Check the vendor’s docs |
| Where you use it | VS Code, JetBrains, Cursor and Windsurf (VS Code forks), CLI, Slack, iOS/Android; web app not yet available | A VS Code fork — and Kilo’s extension also runs inside it | Check the vendor’s docs | Check the vendor’s docs |
| Team plan | Teams at $15/user/month; Enterprise custom | Check the vendor’s pricing page | Check the vendor’s pricing page | Check the vendor’s pricing page |
| Best for | Beginners who want a genuine free start and maximum flexibility | Check the vendor’s docs | Check the vendor’s docs | Check the vendor’s docs |
A few honest notes on reading this table. Competitor cells are deliberately qualitative — pricing and free-tier details for other tools change often, so we point you to each vendor’s pricing page rather than quoting numbers we can’t verify here. The one cross-tool fact from Kilo’s own docs: Kilo’s extension also runs inside Cursor, Windsurf, and other VS Code forks, so trying Kilo doesn’t require abandoning an editor you already like (getting-started docs).
As a rough decision rule: if you want to spend nothing while you learn, start with Kilo Code on Auto Free.
What beginner mistakes cost money or time?
The costliest beginner mistakes are all avoidable: sending confidential code through Auto Free, assuming autocomplete is free, forgetting the 5% credit-purchase fee, and expecting today’s free model lineup to last forever (using Kilo for free). Each mistake below comes with the reason it happens and the fix.
Mistake 1: sending confidential code to Auto Free. This is the big one. Auto Free may route your requests to providers that log prompts and outputs and use them to improve their services — in particular, it may route to NVIDIA’s free endpoints, where trial use is explicitly logged for security and product-improvement purposes. Kilo’s own docs warn: do not submit personal or confidential data when using Auto Free. If you’re working on proprietary code, anything with secrets or keys, or anything covered by an NDA, use BYOK with a provider you trust, or a local model via Ollama or LM Studio instead. Free is a great price until your private code ends up in someone else’s logs.
Mistake 2: assuming autocomplete is free. Autocomplete — those inline suggestions as you type — routes through Kilo’s provider by default and uses credits. If you run out of credits without a free alternative configured, autocomplete simply stops working (your main agent workflow is unaffected). One option is to add your own Mistral API key via BYOK, in which case Codestral autocomplete runs through your Mistral account and its limits instead of Kilo credits — though whether that’s actually free depends on your Mistral plan. Or simply disable autocomplete if you don’t need it.
Mistake 3: forgetting the 5% credit fee. Every credit purchase carries a 5% payment-processing fee. It’s easy to overlook when topping up in a hurry, and it slightly changes the math when you’re comparing pay-as-you-go credits against a Kilo Pass subscription. Budget for it.
Mistake 4: expecting the free model lineup to stay the same. The free models available through Auto Free and the model picker change over time as Kilo partners with different inference providers, and the models behind Auto tiers are updated server-side too. A free model you rely on today can change over time. Check kilo.ai/models for current assignments, type free in the model picker to see what’s live right now, and don’t build a workflow that depends on one specific free model existing forever.
FAQ
The individual platform plan costs $0 (Kilo pricing), and Auto Free routes requests to free models at $0 on Kilo’s side, limited to 200 requests per hour per IP (rate limits and costs). The three questions below cover free use, changing free models, and when to upgrade.
Is Kilo Code really free, or is there a catch?
It’s really free to start, with one honest caveat. The platform plan is $0, Auto Free routes your requests to free models at $0 on Kilo’s side, and free usage is limited to 200 requests per hour per IP (rate limits and costs docs). The catch isn’t a hidden paywall — it’s data handling: free routes may go through providers that log prompts and outputs, so keep confidential code on BYOK or local models instead.
Do the free models ever change or disappear?
Yes. Available free models change over time as Kilo partners with different inference providers, and the models behind the Auto tiers are updated server-side as better options become available or providers change pricing (using Kilo for free). To see what’s currently free, type free in the model picker — free models are clearly labeled. If a free model rate-limits you, just switch to another one labeled “(free).”
When should I upgrade to Kilo Pass or Teams?
Upgrade when your usage pattern demands it, not before. If you’re hitting Auto Free’s hourly limits regularly or want the bonus credits that come with a subscription, Kilo Pass from $19/mo is built for regular usage (Kilo pricing). If you need team collaboration — shared agents, centralized billing, usage analytics — Teams at $15/user/month is the fit. For governance needs like SSO, audit logs, and SLA commitments, talk to Kilo about Enterprise, which is custom-priced.
Where to go next
You now know enough to install Kilo Code, run your first free task, and avoid the classic beginner money traps. When you’re ready to go deeper, start with our full Kilo Code review for a critical look at strengths and weaknesses beyond the official docs. To see how Kilo stacks up against the rest of the market, browse our AI Coding Assistant comparison hub, which collects head-to-head breakdowns. And if you’re curious about the alternatives, our Cursor beginner’s guide covers Cursor from the same beginner-friendly angle. Finally, bookmark Kilo’s own documentation — for a fast-moving tool, the official docs are the fastest way to catch changes as they happen.
📖 Related Reads
- ToolBrain — tool reviews, LLM comparisons, and AI workflow guides
- Hermes Tutorials — Hermes Agent setup, configuration, and advanced workflows
Cross-links automatically generated from None.
Back to all posts