Everything you need,in one source-grounded workspace.

Chat modes, agentic workflows, downloadable files, citation discipline, models, workspaces.

The document workspace: browse documents, notes, and labels, then open a selection in a Focus chat

Three chat modes, one engine

Ask docAnalyzer plans the next step. Focus chat performs deep analysis on a bounded dataset. Co-work chat turns that analysis into working notes, drafts, and edits inside a collaborative canvas. The same chat engine works seamlessly to pass knowledge everywhere inside one session, eliminating the need to switch tabs or paste context between tools.

Ask docAnalyzer: workspace-aware orchestrator that plans, dispatches, and answers product-knowledge questions inline
Focus chat: chat scoped to a chosen set of documents, notes, and labels; every claim cited back to the page
Co-work chat: note-bound chat that edits the note alongside you in a rich-text canvas editor, with citations attached as you write

How chat modes work

Five Agentic Workflows

Batch operations that run the same task across every source in a dataset and return a structured result. More powerful than chat, lighter than custom code. One pass per source, dispatched in parallel, results stitched into one output.

Summarizer: Generate consistent summaries across every source. Configure length, format, and tone to fit your workflow.
Data Extractor: Extract the desired fields from your source and compile them into structured, reusable context you can export in any format.
Individual: Ask the same question to every source and receive source-specific answers, all in one view.
Blueprint: Evaluate every source against a reference document, template, rubric, or checklist. Identify gaps, deviations, and compliance issues at scale.
Humanizer: Rewrite AI-generated text to sound more natural and human while preserving facts, structure, and citations.

How workflows work

Build files, not just answers

When an answer needs to be a PDF, spreadsheet, chart, or diagram, docAnalyzer generates the file natively, and it appears as a downloadable chip in the chat. Later turns can reference, convert, or bundle that chip without re-running the analysis.

PDF, XLSX, HTML, CSV: built inline by the model, downloadable as chips
Vega-Lite charts, Mermaid diagrams, Graphviz, LaTeX math: rendered, not described
ZIP bundles that stitch multiple outputs together
Cross-turn reuse: turn 3's spreadsheet becomes turn 7's PDF without recreating it
Format conversion via Pandoc, LibreOffice, and Calibre under the hood

How downloads work

Citation discipline

Every cited claim is a clickable reference that opens the document viewer at the exact page or section. The multi-round agentic engine retrieves, reads, and refines instead of guessing in one shot.

Clickable references: one click opens the source at the exact page or section
Context-adherence control: a dial you operate, where high keeps the model strictly inside your sources and low lets it draw on general knowledge
Multi-round agentic loop: retrieve, read, refine instead of one-shot guessing
Multi-document citations carry a source prefix so each claim's origin stays traceable
Cited cells in structured outputs: every extracted field links back to where it came from

How citations work

All the models, one plan

30+ models across a dozen providers, in one interface. Switch mid-thread without restarting the agent or losing the active work. If one model is busy or unavailable, the conversation keeps moving on the next one in line.

30+ models across Anthropic, OpenAI, Gemini, DeepSeek, Mistral, Qwen, Groq, and more
Switch model mid-thread without restarting the agent or losing the active work
Resilience: if one model is busy, the conversation keeps moving on the next
Catalog-curated, where gating decides which models appear in which surface
Per-call switching keeps the agent loop intact across model changes

How model selection works

Workspaces, sharing, integrations

Per-tenant isolation by architecture. Documents, notes, labels, and chats sit inside workspaces you own. Bring your existing library in via Zotero or Mendeley. Share chat threads as public links or embed branded Chatbots on your site.

Per-tenant isolation: your content never pools across customers, never trains models
Workspaces group documents, notes, labels, and threads by project
Zotero and Mendeley OAuth import for existing reference libraries
Share chat threads as public links with optional expiration or password
Embed branded Chatbots on your site, backed by a label of documents
Smart Search & Selection: describe what you want in plain English; docAnalyzer pulls matching documents into your working set

How workspaces work