Engineering Blog

Atlarix used to give you one switch: deep thinking on or off. But deep thinking means something different to every model, and a boolean couldn't say any of those things. /effort replaces it with each model's own levels, picked per chat.

You pick a level, but what goes in the request? Behind /effort is a per-model resolution step: the level is matched against what that model declares it accepts, then sent under a camelCase key that one SDK silently drops if you get it wrong. Here's the whole path.

One Atlarix session can message another, set its mode and model, review its plan, and get notified when it finishes. Here is how peer messaging works across the desktop app, CLI, VS Code and any ACP editor — and why it matters for multi-agent workflows.

One Atlarix session can review another's pending changes and plans — approve, deny or replace them — before anything lands. Here is how cross-session review works, why it catches mistakes a single session misses, and how to set it up.

Fork a session, send it a task, set its mode and model, and get notified when it finishes. Here is how to delegate work across multiple Atlarix sessions — the commands, the workflow, and what to watch for.

Atlarix 15.11 can see your screen and use the mouse and keyboard, for the software that has no API. It works with any model that accepts images, is off until you switch it on per model, and stops the moment you move the mouse.

In Atlarix, multi-agent work is built from ordinary chat sessions. Fork a session, let one lead, and it can list the others, send them tasks, set their mode and model, read their results and approve their plans, while you open, read or stop any of them.

Every request a coding agent sends carries its system prompt and tool list. For Atlarix 15.11 we removed what nobody used: the mobile tools, image tools for models that can't see, and the onboarding. Measured on the real request: 452 tokens and two tools fewer per call.

Run Claude as your coding agent's model however your company buys it: your own Anthropic API key, Amazon Bedrock's Anthropic endpoint, or a gateway that speaks Anthropic's Messages API. How to connect each in Atlarix, what to type where, and what works the same on every route.

Atlarix marks its system prompt for the model's prompt cache, and inside one message the cache worked. Across messages it never did: each new message in Build mode wrote about 19,000 tokens to the cache instead of reading them. One block of text was in the wrong place. Here is how we found it, the numbers before and after, and the one-line rule it comes down to.

Atlarix now runs inside your editor. One extension for VS Code, Cursor, Kiro and Windsurf carries the whole agent: every mode, any model or your own API key, edits in the editor's own diff, plans that tick off as they are built, and your Chrome. Zed and JetBrains IDEs get the same agent over the Agent Client Protocol.

Bring your own key from OpenAI, Anthropic, Google, DeepSeek, OpenRouter or another provider, and run a full coding agent in VS Code, Cursor, Kiro or Windsurf on it. Step by step with the Atlarix extension, what your key costs you, and where it is kept.