Ask once, keep the answer.
Threads that read the web, search your own documents and hand back an artifact you can keep.
The thinking is on your keys. The legwork is metered.
Every token the model reads or writes is billed by your provider, never by us. Credits only move when nemu goes out and does something: a search, a page fetched, a research run, a document embedded.
A thread that can actually go and look.
Most chat is a text box in front of a model. This one has hands: it fetches, reads, searches what you have indexed, and writes the result somewhere you can use it.
Any model you mapped
The names you set on the platform show up here. Switch mid thread without losing the conversation.
Attachments that get read
Images, PDFs and documents are parsed to clean text before the model ever sees them.
Web search and reading
Deterministic fetching and conversion, so a page becomes Markdown the same way every time.
Deep research
A run fans out across searches and pages, then hands back a brief with its sources attached.
Artifacts
Long answers become a document you can edit and keep, not a wall of chat you have to scroll.
Connectors and indexes
Point a thread at GitHub, Drive, Slack or your own indexed documents, inside the scopes you approve.
Start with your keys, keep the whole bill.
The gateway, the chat and the computer on one account. Free is enough to evaluate the gateway and the chat on your own keys, and your provider keeps billing the tokens exactly as it does today.