There’s an old public service announcement from a generation ago: It’s 10 p.m. Do you know where your children are?

The question was designed to make people stop and reckon with something they’d been assuming was fine.

Every CIO, CTO, and CISO in the world should be asking a version of that question about AI right now: Do you know where your data is?

Not according to your company’s internal AI policies. Actually, in practice. Builders are running AI agents today on personal accounts, with their own API keys, across five different tools, routing an organization’s most sensitive context through external services no one is monitoring.

For most enterprises, the honest answer is: no.

Today, there is a solution. Anaconda is acquiring Kilo Code.

The Tokenpocalypse Is Here

The global AI economy now processes hundreds of trillions of tokens every month, according to OpenRouter. Kilo alone orchestrates almost 10 trillion tokens per month across more than 3 million developers. These AI-native development workflows are operating at a scale that would have been unimaginable 12 months ago.

Meanwhile, enterprise AI spend is growing faster than anyone’s ability to account for the spend. AI builders are being told to move fast, and the token consumption accumulates invisibly, spread across dozens of tools, work accounts, and personal accounts. It leaves no single view of where it all goes.

There’s been no real answer to token-maxxing. Organizations are spending enormous sums, with very little visibility into where the risk is and if they are getting an ROI.

The harder question underneath the spend is one of dependency. Would any organization accept a single source dependency on anything else this critical to its business? If a cloud provider had downtime, there would be a failover plan that included an alternate cloud provider. If there is a chance your company’s law firm could have conflicts on the business you need from them, you would have another law firm on retainer or ready to go.

Yet many enterprise development workflows today are entirely dependent on a single model provider. If that provider has an outage, changes their pricing, including removing capabilities from your plan, or decides to retire a model suddenly, work stops.

Developers aren’t going to stop using AI. The question is whether enterprises are going to own that experience end to end or just hope for the best.

“There’s been no real answer to token-maxxing. Organizations are spending enormous sums, with very little visibility into where the risk is and if they are getting an ROI.” —David DeSanto, CEO, Anaconda

The False Trade-Off

In conversations with enterprise leaders, the pattern is consistent. CIOs and CTOs are not afraid of AI. They are afraid they’ve lost visibility into what it’s costing, whether their data is being handled securely, and whether they have any real alternative. They feel trapped by what feels like their only option. They have no consolidated view of what their teams are running, and they cannot assess whether the speed they are gaining is commensurate with the risk accumulating underneath it.

The instinct is to frame this as a forced choice: velocity or trust. Move fast or move safely. Give AI builders the tools they love, or maintain the governance posture the board requires. That is a false trade-off. And the fact that so many enterprise leaders believe they must make it is precisely the problem this acquisition is designed to solve.

Right now, most enterprises are living with one of two realities. Either they lock everything down to one tool and one model provider, and the CTO becomes the person holding back the team. Or, they look the other way while developers use whatever they want, with zero visibility. Neither of those is a real strategy.

What enterprises actually need is a path between those two options: any IDE, any model, any provider, with a single layer to see all of it, enforce policy across it, and trust that AI is operating on the organization’s terms.

“Developers aren’t going to stop using AI. The question is whether enterprises are going to own that experience end to end or just hope for the best.” —Scott Breitenother, CEO and Co-Founder, Kilo Code

What We’re Building Together

Anaconda and Kilo have been working on the same underlying problem from different sides.

Anaconda has spent over a decade earning trust at the foundation layer: the packages, the environments, the models, the orchestration layer, and the governance that over 52 million developers and 95% of the Fortune 500 begin their AI work. The principle has always been the same: make sure builders start with the most trusted foundation on the market, and make sure what they build can move from experimentation into production with confidence. That has meant securing the software supply chain, curating model catalogs, and providing the reproducibility and auditability that regulated industries require.

Kilo has been building at the agentic engineering layer, where builders and AI agents actually do the work: optimizing how they build, where they build, what they build with, and how to deploy into production. A model gateway that routes intelligently across 500+ models, including open weight and frontier AI options. Agent orchestration that enables multiple agents to collaborate on a single task, each matched to the right model for the job. Analytics that give organizations one consolidated view of AI development activity across the enterprise.

Anaconda brings the trusted packages and models used by AI builders. Kilo brings the most powerful AI agents to the places AI builders already work. Together, they close the gap between AI that builders want to use and AI that enterprises can trust deploying.

Early results from customers running AI agents through the Anaconda Platform are reporting 30 to 50% reductions in token consumption. Intelligent routing through the Anaconda MCP server ensures agents get the context they need correctly on the first pass, rather than burning tokens through correction cycles. That is governed and intelligent agent routing working at the same time.

AI on Your Own Terms

There are two directions the AI industry can take from here. One leads to a world where a small number of model providers tightly control how AI is accessed, what it costs, and what organizations are permitted to do with it. Tool choice and model choice collapse into a single vendor decision, made by single entities whose incentives are not aligned with the needs enterprises have and how they depend on it. This leads to vendor lock-in, including escalating costs, lack of flexibility, vulnerability to outages, and security risks.

The other path is one where AI is genuinely open. Where the best model for the job is the model the organization chooses, not the model the AI lab prefers. Where frontier cloud models handle complex reasoning, efficient open weight models run routine development tasks, and self-hosted or air-gapped models serve workloads that cannot leave organizational boundaries, all managed from a single platform, on the organization’s terms.

Anaconda has never been in the business of telling customers which cloud to use or which model to trust. We’re flexible, and anti–lock-in, and that’s not changing. We want enterprises to use the best model for what they’re trying to do.

That commitment shows up most directly in open source: Kilo’s open source and source-available codebase and Anaconda’s championing of open source since its start.

Open source is a key component of “AI on your own terms.” There’s a version of the future where AI, the most powerful technology in a generation, is actually available to all, because the platform underneath it is open and agnostic. That’s the world we’re building toward.

What This Means for You

We hear the same tension constantly: AI builders want the freedom to use any model, in any tool, without waiting on permission, and their leaders want to be able to enable them without losing sight of what’s running or what it’s costing. That’s the gap this acquisition closes.

For AI builders: The tools (VS Code, CLI, JetBrains) and environments already in use, are now backed by the governance layer organizations need in order to say yes to them. Kilo remains available for individual builders, teams, and organizations. What changes is that it now sits inside a platform that makes enterprise adoption faster. Build with any model, in any IDE, on trusted infrastructure.

For enterprise leaders: The choice between developer experience and organizational control is not one to make. One platform, one place to see AI development activity across the organization. Intelligent model routing to manage and optimize spend. Policy enforcement that travels with workloads regardless of where they run. If a model is off-limits for a team, it disappears from their selection list. If a provider updates their API, the gateway handles the transition. Governance becomes the blueprint, not the bottleneck.

The Platform Continues

Our acquisition is one chapter in a deliberate sequence. Anaconda’s platform is being built to cover the full AI-native development lifecycle: the governed foundation where AI development begins, the production orchestration layer that makes AI-native workloads run reliably at scale, and now the agentic engineering layer where AI builders and AI agents do the work.

AI that is capable but ungoverned is a liability. AI that is governed but slows development becomes irrelevant.

The only platform worth deploying at enterprise scale is one that is fast, capable, governed, and above all, trusted.

Contact our team to learn how to build safely and securely with the Anaconda Platform.

David DeSanto is CEO of Anaconda. Scott Breitenother is CEO and co-founder of Kilo.