Files, tools and output

Most assistants stop at describing what you should make. Skyber is built to hand back the artifact.

Reading what you upload

PDFs and documents — text is extracted and read in full; image-only pages are routed through OCR.
Photos and scans — handwriting, diagrams, screenshots and whiteboards are read through the multimodal path.
Spreadsheets and data — tabular files are parsed so the model can compute over them, not just describe them.
Attachments live in a private bucket and are served through short-lived signed links.

Tools

Live web search brings current sources into a turn. Agent tools let the workspace act across your conversations, memories and personas rather than only replying. Skyber also exposes an MCP endpoint, so a compatible agent client can list conversations, search messages and read or write memories with your authorization.

Output

Ask for a document and you get a download: PDF reports, spreadsheet workbooks, slide decks, CSV extracts, generated images and rendered charts. Files appear as download cards in the thread and stay attached to the conversation for later.