Anaconda Acquires Kilo Code: Open-Source AI Coding Agent Joins the Enterprise Stack

Anaconda has acquired Kilo Code, the open-source model-agnostic AI coding agent used by over 3 million developers. The deal, announced on July 15, 2026, brings together Kilo’s agentic engineering front-end with Anaconda’s package security, model governance, and production orchestration — creating the first end-to-end pipeline from first prompt to production deployment. Financial terms were not disclosed.

The Scale of Kilo Code

Kilo Code is not a small project. The platform orchestrates approximately 10 trillion tokens per month across more than 500 AI models. It operates in the tools developers already use — VS Code, CLI, JetBrains — as an embedded coding agent that is model-agnostic by design.

Combined with Anaconda’s existing ecosystem of 52 million developers and 95% Fortune 500 adoption, the acquisition creates the largest enterprise AI development footprint outside the big cloud vendors.

The “Tokenpocalypse” — Why Enterprises Are Losing Control of AI Spend

The acquisition addresses a pain point that every large organization using AI tools is beginning to feel: loss of visibility into AI spend, risk, and compliance.

“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

Without a governed pipeline, developers sign up for AI coding tools with personal accounts and API keys, bypassing enterprise security entirely. The fragmentation makes it nearly impossible to answer basic questions: which models are teams using? Are they compliant with data residency policies? Is sensitive code being sent to third-party inference providers?

The 80% failure rate of enterprise AI projects — cited by Anaconda and corroborated in industry reports — maps directly to this governance vacuum. Teams move fast with AI, then hit compliance, security, or cost walls when trying to reach production.

Three Tiers, One Pipeline

The combined architecture has three layers:

1. Kilo Code — the agentic engineering layer: front-end AI agents embedded in developer workflows
2. Anaconda Core — package security, model governance, and software supply chain management
3. Outerbounds / Metaflow — production orchestration (Anaconda acquired Outerbounds in April 2026)

Together, they form a governed pipeline that lets developers use any model (local, cloud, or private) while giving enterprise teams visibility, policy enforcement, and cost control.

What Changes for Existing Kilo Users?

For the 3 million developers already using Kilo, the immediate answer is: nothing. Anaconda has stated that Kilo’s products, pricing, and support remain unchanged for individual builders, teams, and organizations.

What does change is the enterprise option. Organizations that need governed agent routing, consolidated billing, and policy enforcement can now use the combined Anaconda platform. Anaconda reports 30–50% token reduction from intelligent routing via their MCP server — a meaningful cost saving at the scale these tools operate.

”AI on Your Own Terms”

The model-agnostic philosophy is central to the acquisition’s positioning. Kilo was built to let developers use any model, any provider, any IDE. Anaconda’s enterprise platform reinforces that — offering a gateway that routes requests to the optimal model for each task rather than locking users into a single provider.

“Kilo has many enthusiastic users doing agentic engineering with the freedom to use any model. Anaconda spent over a decade building enterprise trust from packages to secure environments. The combination complements each other very well.” — Sid Sijbrandij, Co-Founder of Kilo Code and Executive Chair at GitLab

The Bigger Picture

The acquisition is not an isolated event. It follows Anaconda’s April 2026 purchase of Outerbounds (the company behind Metaflow), forming a deliberate platform strategy. Anaconda is assembling a full-stack AI platform that starts with a developer’s first prompt and ends with a production deployment — governed, observable, and auditable at every step.

For context on the broader shift toward open-weight local coding agents that Kilo’s ecosystem is part of, see Local Coding Agents Go Mainstream. The tool-use patterns that power these agents — including the MCP protocol that Anaconda’s routing server builds on — are covered in Tool-Use Patterns for AI Agents in 2026.

The acquisition resolves a tension that has defined enterprise AI development since the arrival of coding agents: choose speed or choose security. Anaconda’s bet is that with the right pipeline, you no longer have to choose.