Kilo

Freemium

Kilo is an open-source AI coding assistant that works where you already code. It runs inside VS Code, JetBrains, or your terminal. You bring your own API keys, use their built-in provider, or run local models for privacy.

The tool switches between specialized modes for different tasks: Code for writing, Plan for architecture, Ask for exploring code without changes, and Debug for fixing problems. You can also swap models mid-task based on cost or complexity.

The trade-off is clear. You get full control over models and pricing, but you manage the setup and configuration yourself.

Open it if you want model choice and cost control. Skip it if you want a polished tool that just works out of the box.

It fits people who want to own their setup and costs.

• Developers who want model flexibility without SaaS lock-in: You choose from 500+ models or bring your own keys. You pay what the provider charges with no markup.
• Teams comfortable managing their own API keys and provider accounts: You set up billing profiles and switch providers as needed. You control where your code goes.
• Users wanting specialized agents for planning vs. coding vs. debugging: The modes match the task instead of one generic chat. You switch when the job changes.
• Teams with privacy needs or on-premises constraints: You can run local models through Ollama or LM Studio. Your code never leaves your machine.
• Solo developers and small teams wanting model experimentation: You test cheap models for simple tasks and expensive ones for hard logic. You learn what works without being stuck.
• Developers willing to invest time in MCP server setup with z.ai: You get multi-step workflows when configured right. The payoff comes after the learning curve.
• Those needing cheaper alternatives to premium models via MiniMax M3: You get comparable output at a fraction of the price. Your token budget stretches further.
• Developers comfortable with debugging and iterative configuration: You fix setup issues and adjust prompts. You trade polish for flexibility.

These are the parts that stay useful after daily use.

• True model flexibility: You bring OpenRouter keys, use the built-in provider, or run local LLMs. You are never stuck with one vendor's models or prices.
• Multi-IDE support: The same agent works in VS Code, JetBrains, CLI, and cloud. You do not rebuild context when you switch tools.
• Specialized agent modes: Code, Plan, Ask, and Debug each handle one job well. You pick the right tool instead of fighting a one-size-fits-all chat.
• Mid-task model switching: You start with a cheap model for scaffolding, then flip to a strong one for complex logic. You spend tokens where they matter.
• Visible cost tracking: Your spend appears at the top of the window. You know what each operation costs before you commit.

The setup takes attention, then it fades into your workflow.

1. Install the extension for your editor or grab the CLI.
2. Connect your API keys, pick the built-in provider, or set up local models.
3. Pick a mode: Ask to explore, Plan to design, Code to build, or Debug to fix.
4. Review diffs before they apply, and switch models if the task gets harder.

Skip it when the overhead outweighs the benefit.

• You need zero configuration: Setup requires managing API keys, model selection, and permissions. Closed tools handle this for you.
• You want predictable flat-rate pricing: Bring-your-own-key means costs scale with usage. Your bill varies with how much you code.
• You need polished stability across versions: Recent releases broke settings migration and added promotional noise. You may need to stay on older versions.
• You want fast, hands-off results: Cheap models can burn tokens on retries. Progress is slower than premium closed tools.