An open-source coding agent as the antidote to AI model lock-in
The Anaconda Kilo acquisition is the purchase of an open source coding agent that connects to hundreds of different AI models so enterprises can use AI-assisted development without being locked into a single vendor or slipping into a governance-free chaos of tools and accounts.
Anaconda, known for governed open-source packages and environments, has acquired Kilo, a popular open source coding agent designed to be model-neutral rather than loyal to any one AI lab. Kilo’s pitch is simple: plug in OpenAI, Anthropic, Google, Mistral, or self-hosted models and switch as pricing and performance change, avoiding AI model vendor lock-in. This is not a minor side bet. Kilo has grown to more than 3 million users in just sixteen months and connects to over 500 AI models, processing trillions of tokens per month across commercial and open-weight models. In other words, developers have already voted for vendor independence with their keyboards. Anaconda is now trying to turn that developer vote into enterprise policy.
The false choice haunting enterprise AI development
Enterprise AI development has been stuck between two bad options: total control that kills experimentation, or a free-for-all that kills governance. Anaconda’s new CEO, David DeSanto, is blunt about it: “Right now, most enterprises pick one of two false options: lock down to a single tool and model provider, or let developers use whatever they want with zero visibility. Neither is a real strategy.” That quote captures why the Anaconda Kilo acquisition matters more than yet another tooling deal.
The backdrop is rising anxiety about AI model vendor lock-in. Industry leaders have warned that closed model providers gain outsized leverage over enterprise data and workflows. Yet the alternative—letting every team wire up whatever agent, model, and account they prefer—produces “tokenmaxxing”: ballooning token usage treated as progress, with no link to value. DeSanto’s view is that scale is not the problem; governing it is. Kilo represents a third path: an open source coding agent that allows freedom of model choice while still being pulled into a governed platform. This is exactly the middle ground enterprises have been missing.
From Python environments to AI-native workflows for millions of developers
Anaconda has spent more than a decade making it safe for enterprises to use open source software inside their walls, building a business around its widely used Python distribution and package manager. That strategy recently expanded into AI with a governed platform for models and AI development, backed by a funding round of more than USD 150 million (approx. RM690 million) that valued the company around USD 1.5 billion (approx. RM6.9 billion). With Kilo, Anaconda is no longer only governing environments and packages; it is stepping directly into where AI-native development starts: the IDE prompt.
Kilo’s open source roots are tangible. It is a fork of Roo Code, an open source AI coding agent for VS Code, and it has amassed more than 26,000 GitHub stars and 3,000 forks. According to one source, Kilo Code grew from zero to more than 3 million users in sixteen months by word of mouth, driven by its open source foundation and model-agnostic approach. Those users work across VS Code, JetBrains, web, and CLI environments, routing trillions of tokens per month. Anaconda, already trusted by 95% of the Fortune 500, now has a direct line into those workflows. This is not just integration; it is a land-and-expand play targeted at where developers live.
Open-source coding agents as the new control layer
Kilo is part of an emerging category: open source coding agents that sit above individual AI models, offering a buffer between developers and model providers. Tools like OpenCode, Cline, and Aider support this layer, giving developers a way to avoid unpredictable token bills and AI model vendor lock-in by abstracting away the underlying providers. Kilo’s core sell is neutrality, and its self-hosted deployment options and automatic model selection have already made it attractive to enterprises that want to control token costs and govern AI-generated code at scale.
What makes Anaconda’s move distinctive is that it ties this open source coding agent directly into a governance stack. The company already governs packages, environments, and AI models for large enterprises; Kilo connects that governed base to the developer’s first AI prompt. The combined platform is meant to create a trusted, auditable path from prompt to production deployment, so enterprises can move quickly without losing control or failing compliance requirements. In effect, open-source coding agents become not only a freedom tool for developers, but also a new control layer for enterprises—a rare win-win in this space.
What the Anaconda–Kilo path means for the future of enterprise AI
In practical terms, little changes for Kilo’s users today: the product remains available with no changes to existing plans or support. Developers can keep working in VS Code, JetBrains, and the CLI as before, but will gain default access to Anaconda’s vetted packages, governed models, and AI orchestration behind the scenes. Over the next 12 months, Anaconda plans to fold Kilo into its AI workspaces and connect it to orchestration and governance tools so projects can move from code to production without switching platforms, under consistent organizational policies.
Strategically, this positions Anaconda to capture enterprises that want AI-assisted development without giving up flexibility or independence. “Our vision is a single, unified platform where builders don’t have to choose between speed and governance,” DeSanto says. With 21 billion downloads and more than 52 million users already on its platform, Anaconda is building a future where open-source coding agents are the standard entry point for AI, not a sidecar. If that vision holds, the single-model era of enterprise AI could end not with a bang, but with a quiet commit inside an IDE powered by Kilo.
