Kilo ads

Kilo planned a backend build on Auto Frontier and implemented it on Auto Efficient. Total cost $1.26. The same build with a single frontier model running every phase cost $2.90, and passed the same 29 of 30 checks. Auto Frontier pairs a reasoning model to planning and debugging, an implementation model to building. Auto Efficient routes each request to the cheapest model proven accurate enough — inside a pool your admins approve, never above it. Book 30 minutes. We will configure that pool against your approved provider list and run the split on a repo your team actually cares about.

Public. Internal. Confidential. Restricted. Your data already has classes. Your coding agents probably do not respect them. Some free routes log prompts and outputs, or use them for service improvement. Right now that data-handling decision is being made by whoever picked a dropdown. Book 30 minutes with an engineer. We will map your four data classes to permitted routes, show SSO and role-based access controlling which models developers can reach, and walk the audit evidence your security team will ask for.

On Kilo's Terminal Bench 2.0 evaluation, Kimi K3 completed 72.8% of tasks at $48.38 per attempt. Claude Opus 5 completed 71.5% at $113.54. One benchmark does not settle model choice. But most engineering orgs cannot break agent spend down by model, project, and user, so the frontier premium gets paid on work that never needed it, for months at a time. Book 30 minutes. We will show you that breakdown against your own usage, and where the premium is going.

The AI coding tool your developers want is stuck in vendor onboarding. Kilo Code is on AWS Marketplace — same governance, same 500+ models, contracted through an agreement your legal team already signed.

Porting Kilo Code from VS Code to JetBrains turned out to be more than a UI exercise. Rather than embedding the existing VS Code webview in a JetBrains panel, we built a dedicated JetBrains plugin in Kotlin/Swing on top of the IntelliJ Platform SDK. That gave us native IDE behavior for the parts developers notice every day: keymaps, focus handling, theming, layout, and file interactions. The first GA release includes the core Kilo workflow: chat, slash commands, file mentions, model selection, agent settings, MCP server configuration, and attachments. If you already use Kilo in VS Code, the experience should feel familiar while fitting naturally into IntelliJ IDEA, WebStorm, PyCharm, GoLand, Rider, and other JetBrains IDEs. A few advanced VS Code-specific features are still being ported, especially the ones that depend on VS Code APIs. But the core workflow is available now on the JetBrains Marketplace. If you use JetBrains IDEs, we’d be glad to hear how it compares with your current AI coding setup. https://lnkd.in/g9MEaeKj

Ten months ago we predicted AI would cost $100k/dev/year. Ramp's latest data puts the top 1% of companies at ~$90k per developer, up 14% in a single month. But the ones actually in trouble aren't the companies at the top of that spending curve. It's the companies that bet their whole workflow on one vendor and are starting to realize what that costs. As Ramp puts it: "The top 1% of firms tend to mix and match, opting to bounce between multiple frontier models and platforms that give them access to cheaper open source models." We launched Kilo Auto Efficient for that exact reason, to help teams route each task to the right model. I wrote more about what that means here: https://lnkd.in/ejCHraQK