Kilo Code review

Kilo Code is an open-source coding agent for VS Code, JetBrains IDEs, the terminal, cloud agents and Slack. Its homepage says it supports more than 500 models, including local models and models accessed with your own provider keys. The source is MIT-licensed, so users can inspect, fork, modify and self-host it.

The product groups work into Code, Plan and Debug agents, and its Auto Model feature routes tasks using Efficient, Frontier or Free strategies. Kilo Gateway offers access to models across 60-plus providers, according to the pricing page.

The key pricing distinction: the Free plan costs $0 per month for the listed IDE and CLI extensions, but usage is billed separately. You can instead bring your own provider key or use local models. Kilo says Gateway charges exact provider rates with no inference markup; that does not mean model use itself is free.

Kilo’s main draw is choice. You can use the same agent across supported editors, the CLI and cloud agents, pick from 500-plus models, or bring your own keys. The MIT license also gives teams more control than a closed coding assistant. The homepage says there is no inference markup, while the pricing page lists Gateway access at provider rates.

The $0 Free plan includes VS Code, JetBrains and CLI extensions, code completion and generation, and use with any AI model. For teams, the pricing page lists a $15-per-user monthly Teams plan with usage analytics, shared custom agents, centralized billing and priority support. There is no annual platform price shown.

That flexibility comes with real-world caveats, but Reddit feedback includes positive reports: one commenter says Kilo reliably detects rate limits from a free-model provider and retries with exponential backoff. Others mention liking its custom agents or using free models for development. These are individual reports, not a guarantee of consistent results.

Kilo’s low platform price is not a full cost estimate. The Free plan says usage is billed separately; Gateway inference is charged at provider rates. Kilo Pass starts at $19 monthly, with up to $26.60 in credits on Starter, $49 with up to $68.60 on Pro, and $199 with up to $278.60 on Expert. Those are maximum credit amounts, not fixed usage limits. Credit purchases carry a 5% processing fee. Cloud compute is extra: the pricing page lists Code Review at $0.33 per hour and Cloud Agent Docker and Small at $0.60 per hour; Standard is $1.20 per hour. Inference is separate, and the published pricing details do not publish a Free-plan usage cap.

Setup also looks less predictable than the feature list suggests. One Reddit thread describes lost settings and provider configuration problems after an update, while other commenters say they stayed on older versions or switched tools. That is anecdotal, but worth weighing if local models or custom providers are central to your workflow. Enterprise pricing is custom, and the $588 annual Kilo Pass figure is not tied to a named tier on the pricing page.

Kilo Code’s Free plan costs $0 for the platform, but usage is billed separately. Teams costs $15 per user/month and adds shared custom agents, analytics, centralized billing and priority support. Enterprise pricing is custom, so there’s no published figure to compare. The pricing details list a 14-day free trial but don’t say which plan it applies to.

For inference, you can bring your own provider keys or use local models at no platform charge. Kilo Gateway charges provider rates with no markup, but model costs are still separate. Kilo Pass is $19, $49 or $199/month; the listed credit values are “up to” $26.60, $68.60 and $278.60 respectively, with up to 50% bonus credits. That’s a sharp price jump from Starter to Pro, then Expert. The page also lists $588/year with a 50% bonus but doesn’t identify the tier.

Watch the extras: credit purchases carry a 5% processing fee. Cloud compute costs $0.33/hour for Code Review, $0.60/hour for Cloud Agent Docker and Small, and $1.20/hour for Standard. Compute is billed per second; inference is separate.

The product information says Kilo Code works across VS Code, JetBrains, the CLI and cloud agents, so teams can use the same agent in different tools. It doesn’t explain how to migrate existing prompts, settings or project history, or whether those items can be exported. Check those details before moving a team workflow over.

Lock-in looks limited if you use the MIT-licensed source, bring your own provider keys, or run local models through Ollama or LM Studio. But using Kilo Gateway, shared agents or cloud features may leave you with configuration to replace if you switch; the published material doesn’t document an export path. Expect some setup work if you need to configure providers or recreate team settings.

We can’t substantiate reasons people leave based on Reddit: no Reddit posts or comments were included in the published material. The clear cost consideration is that inference and cloud compute can be billed separately from the platform plan. Teams that want one bundled, predictable bill—or published Enterprise pricing—may prefer another option.