Kilo Code Review: An Ambitious Multi-Model Orchestrator for VS Code and JetBrains
10 min read
Updated
Is Kilo Code the right AI coding agent for your workflow? Discover its features, orchestrator mode, zero-markup model pricing, and how it compares.
The AI coding agent space is growing rapidly. Following the success of autonomous agents inside VS Code, a variety of forks and competitors have emerged. Kilo Code is one of the more ambitious options: an open-source agent that began as a fork of both Cline and Roo Code, raised $8 million in seed funding, and claims a user base of over 1.5 million developers.
But does it deliver unique value compared to the original projects? We evaluated the extension, the CLI, and the pricing model to help you decide if it fits your development needs.
Kilo Code operates on its primary domain at kilo.ai. Notable updates to the platform include Auto Model routing, which automatically selects the most appropriate model per subtask (planning, coding, debugging, or explanation) based on the task type and your chosen budget tier. Additionally, its Code Reviewer mode can adapt to specific project conventions via a local REVIEWS.md file (similar to how its Memory Bank context feature works), and the platform supports recent models like Qwen3, GLM, and MiniMax M2.7 alongside its existing library of over 500 options.
What is Kilo Code?
Kilo Code is an open-source AI coding agent that runs inside VS Code, JetBrains IDEs, and as a standalone CLI tool. It takes natural language instructions and executes multi-step development tasks, such as creating files, editing code, running terminal commands, and testing in a browser.
What separates it from the tools it forked is its overall scope. While Cline focuses on being a dedicated VS Code agent and Roo Code emphasizes custom agent modes, Kilo Code aims to be an all-in-one agentic engineering platform. This translates to multi-agent orchestration, cloud execution, code review automation, voice commands, and cross-platform session synchronization.
The core extension codebase is open-source and licensed under Apache-2.0, while the CLI uses the MIT license.
Core Features
Orchestrator Mode
Orchestrator mode is Kilo Code's headline feature. Instead of running a single agent that handles tasks sequentially, this mode breaks a complex task into subtasks and routes each one to a specialized mode: architect for planning, coder for implementation, and debugger for fixing issues.
For example, if you input a high-level instruction like "refactor the authentication module to use JWT and add refresh token support," the orchestrator creates a plan, delegates file changes to the coding agent, routes testing to the debug agent, and coordinates the final results. This multi-agent coordination helps the platform handle complex tasks that might otherwise trip up a single-pass agent.
500+ AI Models with Zero Markup
Kilo Code connects to over 500 AI models through OpenRouter and direct provider integrations. Models like Claude 4.5 Sonnet, GPT-5, and Gemini 3 Pro are available at exact provider rates. Kilo Code does not add a markup or per-token commission, meaning a dollar of Kilo credits translates directly to a dollar of model costs.
You can also bring your own API keys from Anthropic, OpenAI, Google, or other compatible providers, or run local models on your own hardware if you prefer to keep your code entirely on-device.
Memory Bank
AI agents often face context limitations between sessions. Kilo Code's Memory Bank addresses this by storing architectural decisions, project conventions, and codebase context in structured Markdown files directly within your repository.
When you start a new session, the agent reads these files to rebuild its understanding of your codebase. While very large projects can still challenge standard context windows, this approach noticeably reduces the need to explain your project from scratch.
Agent Modes
Beyond the main orchestrator, Kilo Code includes several specialist modes:
- Code mode: Generate, complete, and transform code.
- Architect mode: System design, planning, and high-level technical decisions.
- Debug mode: Diagnose bugs and propose fixes.
- Ask mode: Get explanations and ask questions without modifying files.
- Custom modes: Build your own specialized modes for security audits, performance reviews, or domain-specific tasks.
Each mode restricts the agent's behavior so it stays focused on the task at hand. The debug agent will not rewrite your architecture, and the architect will not edit files directly.
Cross-Platform Support
Unlike editors or plugins that are limited to a single IDE, Kilo Code works across multiple environments:
- VS Code and VS Code-based editors.
- JetBrains IDEs (IntelliJ, WebStorm, PyCharm, etc.).
- CLI for terminal-based workflows.
- Slack integration for triggering agents from chat.
You can start a task on your phone via Slack, pick it up in VS Code, and finish it in your terminal.
Additional Capabilities
- Voice commands: Speak instructions directly to the agent instead of typing.
- Cloud agents: Offload agent execution to the cloud so your local machine stays free.
- Automated code review: Agents analyze pull requests and provide feedback before merging.
- MCP Server Marketplace: Browse and install Model Context Protocol servers to extend the agent's capabilities.
- Inline autocomplete: Standard AI-powered code completions alongside the agentic features.
Kilo Code Pricing
Free Tier
The extension and CLI are free to install and use. If you bring your own API key, you pay the model provider directly and Kilo charges nothing extra. New users also receive $20 in free credits to try out models without setting up API keys.
Kilo Pass (Subscription Plans)
Kilo Pass is an optional credit subscription that lets you buy AI model credits at a discount:
| Plan | Price | Credits | Bonus | Best For |
|---|---|---|---|---|
| Starter | $19/mo | Base allocation | 50% bonus credits | Hobbyists, light use |
| Pro | $49/mo | Higher allocation | Larger bonus | Daily development |
| Expert | $199/mo | Maximum allocation | Largest bonus | Power users, teams |
Purchased credits do not expire, though monthly bonus credits expire at the end of each billing cycle. This pass offers the convenience of managing a single balance instead of multiple provider accounts.
Teams Plan
For organizations, Kilo offers a Teams plan at $15/user/month featuring shared workspaces, admin controls, and consolidated billing across team members.
KiloClaw: Managed Hosting
KiloClaw is a managed deployment service designed to run autonomous coding tasks in the cloud without tying up your local machine. It costs $55/month (with a 1-day free trial and no long-term contract), and AI inference is billed separately. KiloClaw includes Telegram, Discord, and Slack connectivity, allowing you to trigger runs and receive results directly in your chat app.
Realistic Monthly Costs
Based on typical usage patterns:
- Light usage (a few agent tasks per day): $10–$20/month
- Moderate usage (daily development): $30–$60/month
- Heavy usage (multi-hour sessions): $80–$200+/month
Cost Considerations
Orchestrator mode uses more tokens than single-agent mode because it coordinates multiple agents. A task that costs $0.50 with a direct agent might cost $1.50 or more through the orchestrator. While the productivity gains can justify the cost, it is helpful to monitor your balance when running complex workflows.
Pros and Cons
What We Like
- Model flexibility: 500+ models with zero markup allows you to experiment freely.
- Orchestrator mode: Very helpful for breaking down complex, multi-step tasks.
- Cross-platform: Support for VS Code, JetBrains, and CLI covers most developers' setups.
- Open source: The Apache-2.0 license provides code transparency.
- Memory Bank: Reduces context loss and repeated explanations between sessions.
- Free to start: Low barrier to entry with $20 in starter credits and the option to use your own API keys.
What Could Be Better
- Feature sprawl: Combining voice commands, cloud agents, Slack integrations, and in-editor features adds significant complexity to a relatively new tool.
- Orchestrator token costs: Multi-agent coordination consumed credits faster than standard single-agent workflows.
- Early-stage features: Features like cloud agents and automated deployment are still maturing.
- Fork complexity: Built on top of Cline and Roo Code, meaning it inherits some architectural decisions and complexity from its predecessors.
- Ecosystem size: The community-driven marketplace for custom modes and MCP servers is still growing compared to older alternatives.
How Kilo Code Compares
Kilo Code vs. Cline
Kilo Code forked Cline and added orchestrator mode, cross-platform support, and the Kilo Pass credit model. Cline remains focused on VS Code and relies exclusively on user-provided API keys.
- Choose Cline if you prefer a mature, focused VS Code plugin with an established community.
- Choose Kilo Code if you want multi-model flexibility, JetBrains support, or orchestrator mode for handling complex, multi-stage projects.
Kilo Code vs. Cursor
Cursor is an AI-first, standalone code editor built as a fork of VS Code. Kilo Code is a plugin designed to run inside your existing editor.
| Feature | Kilo Code | Cursor |
|---|---|---|
| Type | VS Code/JetBrains plugin | Standalone IDE (VS Code fork) |
| Model Access | 500+ models, zero markup | Multiple models, built-in |
| Pricing | Free (BYOK) or $19–$199/mo | Free / $20/mo / $40/mo |
| Orchestration | Multi-agent orchestrator | Single-agent architecture (Composer) |
| IDE Flexibility | Works in your current editor | Must use the Cursor IDE |
| Code Completion | Basic inline suggestions | Advanced, context-aware autocompletion |
| Best For | Multi-model flexibility & JetBrains users | Polished, all-in-one IDE experience |
- Choose Cursor if you want a highly polished, deeply integrated editor with native AI features and agent capabilities.
- Choose Kilo Code if you want to keep your existing editor setup (especially in JetBrains) and want agentic tools without changing IDEs.
Kilo Code vs. GitHub Copilot
GitHub Copilot is primarily an inline autocompletion assistant with emerging agent features, while Kilo Code is primarily an autonomous agent that also offers basic autocomplete.
- Choose Copilot if you want seamless GitHub integration, predictable subscription pricing ($10–$19/month), and robust inline completions.
- Choose Kilo Code if you want autonomous, multi-file workflows, complete model freedom, and open-source transparency.
Who Should Use Kilo Code?
Best For
- Developers using multiple IDEs: The cross-platform compatibility (VS Code, JetBrains, and CLI) is a major benefit.
- Cost-conscious developers: The ability to bring your own API keys allows you to pay exact provider rates without markup.
- Open-source advocates: Those who want to inspect, self-host, or modify their AI coding tools.
- Developers tackling complex tasks: The orchestrator mode is ideal for managing projects that require step-by-step planning and execution.
Not Ideal For
- Developers looking for maximum stability: Kilo Code's feature set is evolving rapidly, and some features are still maturing.
- Beginners: The wide variety of models, agent modes, and configuration options can be overwhelming.
- Strict corporate environments: Some cloud-based features require sending code snippet context to external servers, which may not comply with all corporate security policies.
Verdict
Kilo Code is an ambitious entry in the autonomous coding space. Its orchestrator mode is a strong differentiator, breaking down complex tasks into coordinated steps across planning, coding, and debugging agents. Combined with access to over 500 models at cost, it offers high flexibility for power users.
While the rapid introduction of features like voice commands and Slack bots creates some complexity, the core agent execution is powerful. If you are looking for an agent that supports JetBrains IDEs or want more control over your model routing and API billing, Kilo Code is a strong option to evaluate.
Our Rating: 7.5/10