Overview

This section highlights the core features, use cases, and supporting notes.

Kilo is an open-source AI coding agent built around one workflow across VS Code, JetBrains, and the CLI rather than a single isolated editor plugin. It is most useful for developers who want current install docs, flexible provider setup, free-or-BYOK model paths, and broader workflow surfaces like cloud agents and code review, while the main tradeoff is that Kilo now behaves more like a real platform with setup choices and cross-interface behavior than like a tiny local-first coding add-on.

Kilo is easier to recommend honestly when we start from the official product pages checked on April 16, 2026 instead of repeating older directory text. The official homepage now calls it the open-source AI coding agent for VS Code, JetBrains, and your CLI. That matters because it immediately corrects the older English page, which felt too generic and too narrow. Kilo is no longer best described as a small local-first helper. The official product story is much broader than that.

Annotated reference image based on the official Kilo homepage showing Kilo positioned as an open-source AI coding agent for VS Code JetBrains and the CLI
The homepage matters because it gives the clearest official positioning in one line: Kilo is meant to span VS Code, JetBrains, and terminal work.

The homepage also makes Kilo’s surface area much clearer than the current English page does. In the Use Kilo Everywhere section, the official site lists VS Code, JetBrains, CLI, Cloud Agents, App Builder, Slack, Code Reviewer, Gateway, and Gastown. That means Kilo should be judged as an engineering workflow platform rather than only as one IDE extension. For readers deciding whether to install it, this is an important correction because it explains why Kilo may feel more ambitious than simpler coding assistants.

Annotated reference image based on the official Kilo homepage showing the Use Kilo Everywhere section across VS Code JetBrains CLI Cloud Agents and other workflow surfaces
The workflow-surface list matters because Kilo is trying to stay with the developer across IDE, terminal, cloud, and collaboration contexts.

The sync story goes even further. The same homepage says users can start a task on a mobile device and finish it in VS Code, JetBrains, or the CLI without missing a beat, and it says session history, active agents, and variables follow automatically across devices and environments. This is one of the strongest reasons the current English page needs rewriting. If the official product promises cross-device continuity and persistent context, then calling Kilo merely local-first undersells the current product shape. Our grounded judgment is that Kilo now leans toward continuity across interfaces, not minimalism inside one editor.

Annotated reference image based on the official Kilo homepage showing cross-interface sync and persistent context claims
The sync story matters because it changes how Kilo should be evaluated: not as one isolated pane, but as a continuing workflow across interfaces.

Provider choice is another area where the official docs are far more concrete than the current page. The Setup & Authentication docs say Kilo Code supports over 30 providers and give examples such as ChatGPT Plus or Pro with no API key, OpenRouter, Anthropic, and OpenAI. The same page also points to broader options including Google Gemini, DeepSeek, Mistral, Ollama, AWS Bedrock, and Google Vertex. That matters because Kilo is not only selling a single model relationship. It is selling a workflow that can sit on top of several billing and infrastructure choices.

Annotated reference image based on the official Kilo authentication docs showing support for more than 30 providers including ChatGPT Plus or Pro OpenRouter Anthropic and OpenAI
The provider docs matter because developers often decide whether to keep a coding tool based on model choice, billing control, and reuse of existing subscriptions.

Kilo’s openness story is also worth describing more carefully than the old page did. The official Open Source Commitment page says Kilo Code is open and highlights Apache-2.0 licensing, while linking the VS Code extension, JetBrains extension, and CLI repositories. It also says features that are open source today stay open source tomorrow, promises continued BYOK support, and says gateway and cloud backend source is available, with abuse-protection systems excluded for security reasons. This is a more nuanced and more trustworthy story than simply calling the whole product open source with no detail.

Annotated reference image based on the official Kilo open commitment page showing Apache 2.0 licensing BYOK support and concrete openness commitments
The open-commitment page matters because it lets readers separate the genuinely open parts of Kilo from the broader hosted platform story.

The installation docs are equally important for a practical recommendation. On the official install page, Kilo says the current VS Code extension is built on the Kilo CLI and distributed as the pre-release version on the VS Code Marketplace. The same docs are careful to say that the pre-release label is only the Marketplace distribution channel and that the extension is stable and recommended for all users. They also document Open VSX and direct VSIX installs for compatible editors. That is a strong reason to replace the thin old download guidance with something current and specific.

Annotated reference image based on the official Kilo installation docs showing the current VS Code extension pre-release channel plus Open VSX and VSIX options
The install docs matter because they prevent readers from misunderstanding the pre-release label and missing the Open VSX or VSIX fallback routes.

The CLI docs are even more straightforward. They recommend installing the CLI with npm install -g @kilocode/cli, warn that older CPUs without AVX support may need a baseline build from GitHub releases, and tell users to run kilo and use /connect for first-time provider setup. This is practical advice that directly reduces setup friction. It also shows that Kilo’s CLI is a real first-class surface, not only a technical side note behind the IDE experience.

Annotated reference image based on the official Kilo CLI docs showing npm installation baseline-build guidance and the /connect setup command
The CLI docs matter because they set better expectations around npm setup, older CPU compatibility, and the first authentication step.

JetBrains users also need a clearer summary than the current page provides. The official JetBrains docs say Kilo Code supports IntelliJ IDEA, WebStorm, PyCharm, and more, but they also list two setup prerequisites that should not be hidden: JetBrains Toolbox is recommended because authentication callbacks work properly with it, and Node.js LTS is required for the extension’s backend services. That does not make JetBrains support weak, but it does mean the page should be honest that the JetBrains path is a little heavier than a casual one-click install.

Annotated reference image based on the official Kilo JetBrains docs showing supported IDEs and the JetBrains Toolbox plus Node.js prerequisites
The JetBrains page matters because it keeps expectations realistic: support is real, but it comes with clearer prerequisites than a tiny autocomplete plugin.

The official free-usage docs add one more important correction. Kilo says it can be used completely free of charge, but the docs are explicit that model inference appears in three places: agentic interactions, autocomplete, and background tasks. They recommend Auto Model Free for agentic use, explain that free autocomplete can be set up with a Mistral Codestral key through BYOK, and say background tasks need the small model switched to a free model if users want zero credit usage there too. This is much more helpful than vague marketing because it tells readers where the real setup work is.

Annotated reference image based on the official Kilo free-usage docs showing the three AI usage paths and the need to configure free models for all of them
The free-usage page matters because it shows that zero-cost use is possible with Kilo, but not fully automatic across every feature.

The release feed gives the final grounding point. On April 16, 2026, the latest official GitHub release endpoint for Kilo-Org/kilocode returned v7.2.10, published on April 15, 2026. The release assets included CLI archives such as kilo-windows-x64.zip and kilo-windows-x64-baseline.zip, plus VSIX packages such as kilo-vscode-win32-x64.vsix. That fits the official product story well: Kilo is maintaining both terminal and editor distribution paths at the same time.

Annotated reference image based on the official Kilo GitHub release page showing v7.2.10 published on April 15 2026 with CLI archives and VSIX packages
The release page matters because it gives users one authoritative upstream version and asset checkpoint before they install or update Kilo.

Our grounded judgment is that Kilo is most worth trying for developers or teams who want one AI coding layer across VS Code, JetBrains, and the CLI, and who care about provider choice, BYOK flexibility, and a broader workflow surface that can extend into cloud agents and review flows. It is less suitable for people who only want a tiny zero-setup autocomplete helper or who specifically want a purely offline and purely local coding assistant with minimal platform behavior. Kilo now has enough surface area that setup choices, provider choices, and interface choices should be treated as part of the product, not as side details.

Setup / Usage Guide

Installation steps, usage guidance, and common notes are maintained here.

The most practical way to test Kilo is to treat it as a cross-interface coding workflow, not as a single editor extension. The official site and docs now present four real decisions up front: which surface you want to start in, which provider path you want to use, whether you care about free usage or BYOK, and whether the broader cloud-review platform features matter to you. A good evaluation should answer all four.

  1. Start from the official product page at https://kilo.ai/code. Read the homepage title and the Use Kilo Everywhere section first so you understand that Kilo is meant to span VS Code, JetBrains, the CLI, and broader workflow surfaces.
  2. Decide which interface you want to evaluate first. If you already live in VS Code or a VS Code-compatible editor, the install docs are the fastest route. If you work mainly in JetBrains, read that platform page before installing. If you prefer terminal-first work, start from the CLI docs.
  3. For VS Code, open https://kilo.ai/docs/getting-started/installing. The docs say the current extension is distributed as the pre-release version on the VS Code Marketplace, and they explicitly say that this channel is stable and recommended. If you use a marketplace-restricted editor, review the Open VSX and direct VSIX options on the same page.
  4. For CLI-first use, install Kilo with npm install -g @kilocode/cli. If you are on older hardware and see an illegal-instruction problem, go straight to the GitHub release page and use the appropriate -baseline archive for your platform.
  5. For JetBrains, read https://kilo.ai/docs/code-with-ai/platforms/jetbrains before you assume setup is trivial. The docs recommend JetBrains Toolbox for authentication callbacks and require Node.js LTS for backend services. Install those first if you want the smoother route.
  6. After installation, do the first authentication step on purpose. The setup docs say Kilo supports over 30 providers, including routes such as ChatGPT Plus or Pro, OpenRouter, Anthropic, OpenAI, and many others. Choose the provider path that matches how you already pay for or manage models.
  7. If your main goal is to use Kilo with no ongoing Kilo spend, do not stop after basic login. Open the official free-usage page and configure all three model-use paths: agentic interactions, autocomplete, and background tasks. The docs say zero-cost use depends on all three, not just the main chat or agent flow.
  8. The easiest free starting point for agentic usage is Auto Model Free. If you also want free autocomplete, the docs point to a Mistral Codestral BYOK route. For background tasks, switch the small model to a free model so session titling and summarization do not consume paid credits.
  9. Run Kilo inside a real repository rather than in an empty folder. Ask it to inspect a code area, explain context, and make one useful change. If cross-interface continuity matters to you, deliberately start in one surface and continue in another so you can judge whether the sync and persistent-context story is genuinely helpful.
  10. If you care about team workflows, review the official product surfaces such as Cloud Agents, Code Reviewer, or related platform features from the homepage. Kilo is broad enough now that these workflow extensions may matter as much as the editor chat itself.
  11. If openness and procurement matter in your environment, read https://kilo.ai/open. That page is the right place to verify the Apache-2.0 components, BYOK commitment, and the more nuanced status of gateway and cloud backend source availability.
  12. Before manual installs or version-sensitive rollouts, verify the latest release assets at https://github.com/Kilo-Org/kilocode/releases/tag/v7.2.10. On April 16, 2026, that was the latest official release, published on April 15, 2026.
  13. After a few real tasks, judge Kilo on one concrete question: is it reducing friction across your IDE, terminal, provider setup, and review workflow, or is it adding platform complexity without enough payoff for your way of working?

A practical Kilo evaluation usually means choosing the right surface first, respecting the documented install and prerequisite notes, selecting a provider path deliberately, configuring all three model-use paths if cost matters, testing one real repository task, and then deciding whether Kilo's broader cross-interface and cloud-workflow story is valuable enough to keep in the stack.

Related Software

Keep exploring similar software and related tools.