Kilo - Kilo: The Open Source AI Coding Agent for VS Code, JetBrains, and your CLI

What

Kilo Code is an open-source AI coding agent for developers that works in VS Code, JetBrains IDEs, and the CLI. It is designed to help write, debug, refactor, review, and deploy code while keeping work in the developer’s existing environment rather than forcing a separate interface.

The product appears positioned as an all-in-one agentic engineering platform for individual developers and teams. Its core workflow centers on persistent context across interfaces, access to a large model catalog with bring-your-own-key support, and multiple task-specific modes and agents for planning, coding, debugging, review, and deployment.

Features

• IDE and CLI coverage: Kilo runs in VS Code-based editors, JetBrains IDEs, and the terminal, which reduces workflow switching across coding environments.
• Persistent sessions across interfaces: Sessions carry agents, context, variables, and progress between devices and environments, making it easier to continue work without re-establishing state.
• Access to 500+ AI models: Developers can choose among frontier, local, custom, and provider-hosted models, with BYOK and a unified gateway for flexible cost and capability tradeoffs.
• Agentic workflow modes: Built-in modes such as Ask, Architect, Code, Debug, Orchestrator, and Custom support different stages of software delivery with more structured AI assistance.
• Automated code review and deployment: Kilo includes AI code review for pull requests and one-click deployment from the product, helping teams move from coding to validation and release in fewer steps.
• Context, memory, and recovery tools: Memory Bank, automatic context search, test-running, failure recovery, and documentation lookup through its MCP marketplace are intended to improve continuity and reduce low-quality outputs.

Helpful Tips

• Check model governance early: With 500+ models and multiple routing options, teams should define which models are approved for which tasks to avoid inconsistent output quality and uncontrolled spend.
• Start with a narrow workflow: Products like this are adopted more effectively when teams begin with one or two use cases such as debugging, code review, or autocomplete before expanding to orchestration and deployment.
• Review openness and inspectability in practice: Kilo emphasizes open source, visible prompts, and visible context windows, so technical buyers should verify how this transparency fits their internal engineering and security review process.
• Use memory and custom modes deliberately: Shared memory and custom workflows can improve team consistency, but they also need ownership and maintenance to stay accurate as architecture and standards evolve.
• Validate cloud-agent fit by task type: Long-running and resource-intensive work may benefit from cloud agents, but teams should still compare those workflows against local IDE usage and CLI-based execution patterns.

OpenClaw Skills

Kilo’s site explicitly references OpenClaw through KiloClaw, described as a personal AI agent powered by OpenClaw, and also highlights “skills” as sharable packages of domain expertise, new capabilities, and repeatable workflows. Based on that, a likely OpenClaw connection is the creation of reusable skills that package engineering playbooks, support procedures, security checks, code review heuristics, and team-specific development conventions for use across Kilo environments.

In practice, this could enable OpenClaw-based agents and skills that support software teams beyond code generation alone. Likely use cases include release-readiness reviewers, onboarding guides that pull from Memory Bank, Slack-based engineering assistants, vulnerability triage workflows, and architecture-aware debugging agents that operate across IDE, CLI, and cloud contexts. If implemented well, that combination could shift engineering work from isolated AI prompts toward durable, shared operational knowledge embedded in repeatable agent workflows.