Introducing Incbury and Atom

Today we are opening Incbury Chat to everyone, with a free plan that includes every reasoning mode, web search with citations and deep research.

Why another assistant

Most AI products rent their intelligence. That is a reasonable way to start, but it means the people building the product do not control what it will say, how your data is treated, or what it costs next year. We wanted a workspace we could trust with our own work, so we built the model too.

Atom is a mixture-of-experts language model. Every token is routed through a small subset of experts, which keeps it quick without giving up depth. We design the architecture, run the training pipeline and serve the checkpoints ourselves.

What you can do today

• Choose how hard Atom thinks. Instant for speed, Medium for a short think, High for extended reasoning with the working shown before the answer.
• Search with receipts. Turn on search and every claim is linked to the page it came from. Ask for deep research and Atom plans queries, reads across pages and writes a structured report.
• Keep it organised. Projects, pins and archive in the sidebar, and search across every conversation with Cmd K.
• Bring your files. Documents, spreadsheets, images and voice notes all work in the same thread.
• Run code safely. When something needs computing, Atom executes Python in an isolated sandbox and shows the output.

Private by default

Conversations are never used to train Atom without your explicit consent. You can export or permanently delete any conversation. Two-factor sign-in is available on every plan, including Free. Our privacy policy lists every sub-processor and where your data is hosted.

What is next

A Pro plan with higher limits is coming soon, followed by team billing. We will publish evaluation results for the current Atom checkpoint on the evaluations page. You can see the rest of our plans on the roadmap.

Thank you for trying it. Tell us what breaks.