Kilo is a central player in the shift from AI-assisted coding to agent-directed engineering. Their primary contribution is an orchestration layer that coordinates multiple specialized agents to complete complex software tasks. By moving beyond the "chat box in an IDE" paradigm, they provide a blueprint for how technical agents can operate with high autonomy while remaining under human supervision.
The company is active at both the developer tool level and the infrastructure level. Through Kilo Code, they provide the interface for agent-human collaboration, while KiloClaw and their model gateway provide the execution environment for persistent, always-on agents. They are significant to the ecosystem as champions of open-source agent logic and usage-based pricing, offering a modular alternative to the vertically integrated, proprietary agent platforms emerging from major cloud providers.
Kilo is a San Francisco-based technology company founded in 2025 that develops tools for what it calls agentic engineering. Rather than treating AI as a simple autocomplete helper, the company builds systems where developers act as managers of specialized AI agents. This approach is codified in Kilo Code, an open-source platform that integrates with VS Code, JetBrains, and the command line. The system utilizes an orchestrator-agent pattern to break down software tasks into manageable steps, allowing for a higher degree of automation than traditional AI assistants.
The platform is built around five distinct agent modes. The Architect Agent designs feature structures, the Orchestrator Agent distributes tasks, the Code Agent writes the implementation, and the Debug Agent handles troubleshooting. This workflow culminates in an AI Code Reviewer that evaluates the work before deployment. This modular structure reflects the company philosophy of "Kilo Speed," which emphasizes reducing the friction of peer reviews and context switching by allowing individual engineers to own features from design to production.
Kilo is founded by Scott Breitenother, formerly of Brooklyn Data Co., and Sid Sijbrandij, the co-founder of GitLab. Sijbrandij's involvement suggests an open-core business model similar to the one that defined GitLab. Kilo Code is distributed under the Apache-2.0 license, a move that positions the company as a transparent alternative to proprietary editors like Cursor. By keeping the core agent logic open, Kilo allows engineering teams to inspect the agent's reasoning and customize the platform for specific internal environments.
This commitment to openness extends to how the platform handles AI models. Kilo provides access to over 500 different models through its gateway, but it also allows users to bring their own API keys. This model-agnostic approach is a hedge against vendor lock-in, enabling developers to switch between providers like OpenAI, Anthropic, or local models based on the specific requirements of a task or the sensitivity of the code being processed.
While coding is the primary focus, the company is expanding the agent concept into general productivity with KiloClaw. This is a 24/7 personal AI agent that runs on hosted infrastructure rather than a local machine. It connects to surfaces like Slack, Discord, and Telegram, where it can manage calendars, read emails, and browse the web on behalf of the user.
KiloClaw represents a broader vision for the company where agents are not just tools for writing software but are active participants in a professional's daily administrative life. By managing the underlying infrastructure and security, the company attempts to lower the barrier for non-technical users or developers who want automation that persists without an open laptop.
The business model for Kilo is built on economic transparency. For individual developers using the IDE extension and CLI, the tool is free. For teams and enterprise users, the company charges for collaboration features like centralized billing and usage analytics, but it provides AI inference at cost. This "no markup" strategy is a distinct pivot from the subscription-heavy models used by most AI tool providers. It aligns the company's incentives with the user's desire for efficiency rather than maximizing token consumption.
Open source AI coding agent for IDEs and CLI.
Kilo is hiring.