Kilo is the fastest growing coding agent because it is open.
#1 on OpenRouter. 500k+ Kilo Coders. 6.1 trillion tokens used per month.
Agent Orchestration
Use the right mode for each task and run agents in parallel
Orchestrator mode
Breaks downs complex projects into manageable subtasks, assigns them to other agent modes, and coordinates between them.
Features
Everything you need for agentic development
Install where you work
Automatic failure recovery
Nothing is more frustrating than telling an AI about the error it created. Kilo automatically detects the errors, runs test suites and recovers on failure.
Hallucination-free code
Kilo's MCP marketplace lets you add tools like Context7 that automatically look up library documentation, so your AI actually follows best practices instead of making stuff up.
Context is king
Kilo searches for the right context automatically. Any context you add is bonus, not required.
Kickstart debugging
Have Kilo's 'debug' mode go through your codebase with a fine-toothed comb to find where a bug is coming from. No longer is it your job to do the most boring thing in programming.
Memory bank
Stop re-explaining your application to the AI, have AI keep its own records of what it worked on, your preferences and more.
Parallel agents
Run multiple AI agents simultaneously to tackle multiple problems at once. Parallel mode lets you dramatically speed up your work on complex projects.
Tab autocomplete
Great inline completion including multiline suggestions. Autocomplete in Kilo is fast and knowledgable.
Coming soon
- Cloud agents
- Code reviews
- Deploy
- App builder
- Code indexing
Open Models
Switch between 500+ custom, local, and frontier LLMs.
Kilo is the best gateway to discover and compare models. Use new models the day they come out and try stealth models for free.
Minimax M2 Growth
Growth of the latest open source model - MiniMax M2 - in Kilo Code. Want your model to take flight? Get in touch
Open Pricing
Run AI where you want and pay as you go
Kilo Gateway, open & transparent
- Choose from more than 60 providers
- Use more than 500 models
- Pay the exact list price of Anthropic, OpenAI, and Google
- No commission or hidden fees, we make money on Teams/Enterprise
- Pay as you go, no subscriptions, no waste.
- No need to wait for a daily reset, no slow pool, no rate limits!
Or use any alternative for AI inference
- Other Gateways (OpenRouter, Vercel, Requestly)
- Managed Services (AWS Bedrock, Azure OpenAI, Google AI Studio)
- Free, stealth, and budget models (Llama, Mistral, etc.)
- Fallback when you hit a quota limit with our virtual provider
- Bring your own API keys (BYOK) for any provider/subscription
- Run models locally (Ollama, LM Studio)
Open Source
Inspect, control, contribute
The complete plugin is open-source and licensed under Apache-2.0. You can see, control, and modify how Kilo works. Kilo does not have an auto model, so you always know exactly what model is being used. No silent context compression or cut-off. You can see the context window size on each request, along with the full prompts. Join our community on Discord and contribute to help shape Kilo's future.
Coming to Kilo from Cursor? You're not alone

As seen in San Francisco
43%
of Kilo Coders previously used Cursor
I switched my org to Kilo from Cursor. It ROCKS!
Christopher Snyder
Kilo Code User
Recent posts
Read the latest news and updates from the Kilo Code team.
Benchmarking GPT-5.1 vs Gemini 3.0 v Opus 4.5 across 3 Coding Tasks
Three AI giants released their best coding models in the same month:
Claude Opus 4.5 & Autocomplete Webinar: Recap
We hosted a Live Session on Claude Opus 4.5 and Kilo Autocomplete. Here’s how it went.
Kilo Code Weekly Product Roundup
Extension versions 4.120.0 - 4.121.1 | CLI version 0.6.0