Kilocode Acquired by Anaconda

Anaconda has acquired Kilocode, an open-source, model-agnostic coding agent used by millions of developers. The acquisition aims to integrate Kilo's flexible AI-assisted coding tools with Anaconda's enterprise platform, offering developers faster, governed AI at scale without vendor lock-in.

Kilo Code, also known as “Kilo,” is an open source, model-agnostic, coding agent that lets you write faster, better software using your favorite IDE or the command line. If you’ve spent any time in VS Code, JetBrains or your terminal writing AI-assisted code over the past year, there’s a decent chance that Kilo has come across your desk and you’ve tried it out. You may be one of the millions of developers who have integrated it into your workflow. It is the most-used client on OpenRouter, it’s open source, and it doesn’t lock you into one model provider or charge a fee as your cloud middleman. This post will explain why I am so excited about what happens when Kilo, a tool I’ve already been using, connects with the Anaconda ecosystem, how the Kilo approach fits with the way that Anaconda is approaching developer experience, and what this means for how you build today and six months from now. I’m just as excited to welcome the Kilo community to the Anaconda family. Acquisition announcements usually come from the top: Read the official announcement from David DeSanto Anaconda CEO and Scott Breitenother Kilo CEO to understand the enterprise case for why this matters: AI at enterprise scale needs to be fast, capable, governed, and trusted, and spend visibility is a big part of getting there.. This post, however, is the developer-focused point of view on why I and millions of developers have chosen Kilo regardless of what company we work for. What Kilo Code actually is If you’re arriving fashionably late to the party, I’ll run down what Kilo is and how you can leverage it today. Kilo Code started in early 2025, co-founded by Scott Breitenother, Emilie Schario, and Sid Sijbrandij GitLab’s former CEO . It grew fast, being genuinely useful, and it doesn’t make you bet on one AI vendor or another before you get value from it. That’s exactly how I first heard about it: word of mouth from my peers. A few things that matter to me in regards to how I pick my toolchain and play into my excitement around this acquisition: It’s open source and MIT-licensed . You can read exactly what it does, fork it, and audit its behavior. It aligns with Anaconda’s history and our open core future. It’s made for builders and lives where you build. Whether you’re reaching for VS Code, JetBrains or your CLI—installable via npm, curl, pnpm, bun, Homebrew, AUR/paru covering major package managers across macOS, Linux and cross-platform tooling—you can continue building within the developer context you prefer. 500+ models, and you pay the provider’s rate; or go local in your IDE. No markup, no lock-in, no “cloud middleman” fee, and you can switch models mid-task. For example, you can use a fast, cheap model for boilerplate code and a heavier reasoning model for the gnarly parts within the same session. Local-first crowd has an answer. If you’re also interested in coding first with models running locally and then expanding to cloud models when you need more compute, Kilo supports this fully in IDE, with more limited support in the CLI today. This matters if you care about keeping code and data on your machine, working offline or just tinkering with local inference the way I do on my own home lab setup. Agent modes FTW. The first thing I explored was the hotly anticipated agent modes when I installed Kilo to my IDE. If you use Architect Mode for planning before you write a line, then Debug Mode for tracing down what broke, you’re pushing the limits beyond what most code assistants will do for you. KiloClaw extends the agentic tooling. A hosted, always-on agent you can reach through other communication vectors like Telegram, Discord, or Slack that keeps working after you’ve closed your laptop, for scheduled jobs and longer-running tasks. No more keeping your laptop propped open as you move about on the fly. If you’re already using conda environments, Anaconda’s packaging ecosystem, or the broader Anaconda Platform, the direction here only enhances your workflow: Kilo is a builder-focused tool to reach for first that connects to the infrastructure you already trust. These features do not change because of an acquisition. In fact, we’re saying plainly that what you love about Kilo is what we love about Kilo and you can continue to depend on the way it delivers efficiency for you. Sensible defaults, Kilo Code, and conda Kilo’s agents start with no permissions and earn access incrementally, the same instinct that’s kept conda environments useful for over a decade. You don’t hand-edit every dependency pin from scratch; you start from a reasonable, reproducible baseline and adjust from there. Sensible defaults are what let you move quickly without babysitting every step, and also build trust in your tools. In practice, you’re iterating fast, trying a refactor, letting the agent touch files, and switching models mid-task. Every action the agent takes gets logged, whether it be files touched, models called, or permissions expanded. Traversing the records your agent created so you can understand the actions taken in pursuit of your coding goals can help you figure out why your test broke. Similarly, conda runs with sensible defaults so you can move fast, reproduce your environment, and not build from scratch every time, all with transparency that “it works on my machine” with actual teeth behind it. We’re operating in a different layer of your developer stack, but with the same practical approach. Getting started with Kilo Code today Nothing changes about your setup. Today, Kilo works exactly as it did last week. Same models, same pricing, same open-source repo, same install process. If you’re already using it, nothing breaks and nothing requires action from you. Kilo remains available and free for individual developers. For new users, Kilo Code has excellent documentation https://kilo.ai/docs/ on setup, model configuration, MCP server integration, and CLI usage so you can go from zero to hero, working with agents to enable better coding experience today. What’s coming: Anaconda brings a governed package and environment ecosystem that tens of millions of users already rely on for trusted and reproducible environments and artifacts; this ecosystem will be connected to the existing features of Kilo. Bringing Kilo’s agent workspace to the layer that Anaconda has been delivering on as pioneers in the AI, machine learning, and data science space means that the gap between builders hacking locally and builders shipping reviewable, reproducible systems gets smaller. We’re here to enable your experimentation where your ideation starts and promote it to hardened, real-world environments when you’re ready. As Kilo continues to integrate further into the Anaconda ecosystem, expect enablement to expand. If you have questions or recommendations on how we may build environment-aware guides, deeper MCP examples, hands-on content, and our continued investment in the open-source project itself, contact the Anaconda team at email protected /cdn-cgi/l/email-protection caa9a5a7a7bfa4a3beb38aaba4aba9a5a4aeabe4a9a5a7 . Try out the tooling if you haven’t already and let us know what you think: npm npm install -g @kilocode/cli curl curl -fsSL https://kilo.ai/cli/install | bash Homebrew macOS / Linux brew install Kilo-Org/tap/kilo Learn more: We’ll be digging into these capabilities along with others in the expanded Anaconda ecosystem on Numerically Speaking LIVE https://www.youtube.com/playlist?list=PLGB9meziqbzoO2AaM-0TWJDmXcNyJK4Jb over the coming months. If there’s something specific you’d like me to share, tell me. This is an open-core tool with a community behind it, and that doesn’t stop now. We’re eager to continue supporting the community and to expand the conversation in partnership with Kilo superfans and contributors.