Anaconda Acquires Kilo Code as Enterprises Seek More Control Over AI Development
Anaconda has acquired Kilo Code, bringing Kilo’s agentic engineering tools into Anaconda’s platform for AI development. The companies claim the combined offering will give enterprises a unified view of model usage, spending, data flows, and policy enforcement across development environments.
Financial terms were not disclosed.
The acquisition comes as organizations rely on AI agents and multiple model providers, often through a mix of company and personal accounts. Kilo’s announcement claims that developers are routing sensitive business context through external services that may not be visible to enterprise security teams.
Managing a growing token bill
The announcement cites OpenRouter data indicating that AI services process hundreds of trillions of tokens each month. Kilo claims that it orchestrates almost 10 trillion tokens monthly across more than 3 million developers.
Those figures point to the scale of AI-assisted development, but they do not by themselves establish how much value organizations receive from that activity. Kilo argues that spending is often scattered across multiple tools, work accounts, personal accounts, and API keys, leaving enterprises without a consolidated view.
The company also points to dependency risks. Development teams that rely on a single model provider can be affected by outages, pricing changes, altered product capabilities, or the retirement of a model. Kilo and Anaconda argue that routing work across multiple providers can reduce exposure to those changes.
A governance layer across tools and models
Anaconda brings its packages, environments, models, orchestration tools, and governance capabilities to the arrangement. The company claims that more than 52 million developers and 95% of Fortune 500 companies use its foundation for AI work.
Kilo contributes tools aimed at the agentic engineering layer. According to the announcement, those include:
- A model gateway that routes requests across more than 500 models
- Agent orchestration for coordinating multiple agents on a task
- Analytics intended to provide an enterprise-wide view of AI development activity
The proposed platform would allow organizations to use different IDEs, models, and cloud providers while applying policy through a shared layer. The companies describe that approach as an alternative to either restricting teams to a single tool or allowing unmanaged experimentation.
Kilo claims that early customers running AI agents through the Anaconda Platform have reduced token consumption by 30% to 50%. The announcement attributes those reductions partly to routing through the Anaconda MCP server, which it claims can provide agents with the required context on an initial request rather than through repeated correction cycles. No independent methodology or customer details were provided.
An open-model position
The companies are also positioning the acquisition against tighter dependence on individual model vendors. The proposed setup would support frontier cloud models, open-weight models, and self-hosted or air-gapped models, according to the announcement.
That flexibility is tied to both companies’ open-source positions. Kilo points to its open-source and source-available codebase, while Anaconda highlights its open-source work.
The arrangement does not eliminate the possibility of platform dependence. Instead, it moves more control into the Anaconda platform, where model selection, routing, analytics, and policy enforcement would be managed. Whether that creates meaningful choice or a different form of lock-in will depend on how broadly the combined platform supports external tools and providers.
What changes for developers and enterprise teams
Kilo will remain available to individual developers, teams, and organizations, according to the announcement. Its tools include integrations for VS Code, CLI, and JetBrains environments.
For enterprise teams, the companies promise centralized visibility into AI development, model routing intended to manage costs, and policies that follow workloads across deployment locations. The announcement also claims that administrators could remove restricted models from a team’s available choices and use the gateway to handle provider API changes.
Anaconda characterizes the acquisition as an extension of its existing platform strategy. The company aims to cover the governed foundation for AI development, the production orchestration layer, and the engineering environments where developers and AI agents build software.
David DeSanto, Anaconda’s CEO, and Scott Breitenother, Kilo’s CEO and co-founder, authored the announcement. Their central argument is that enterprise AI adoption requires both development speed and stronger oversight. The acquisition gives Anaconda a broader set of tools to pursue that goal, while its cost and governance claims remain largely based on company-provided figures.
Source: Kilo Blog
