The OrchestratorOverviewHow It Connects
The StudioOverviewExtensionsHow It Works
ResourcesBlogFAQAbout
Get in touch
Early Access
Studio Extension

Documents

PDF, DOCX, XLSX, PPTX — read, write, convert

Your agents need to read a contract, extract data from a spreadsheet, or generate a report. The Documents extension processes files locally — no cloud upload, no third-party API, no file contents leaving your machine unless you choose to.

Drag and drop with conversion
Drag and drop with conversion
Converted document with LLM interpretation
Converted document with LLM interpretation
PDF to Markdown conversion with annotation
PDF to Markdown conversion with annotation
Document management with PowerPoint preview
Document management with PowerPoint preview

Drop any file into the chat — the plugin detects the format and converts it automatically.

Capabilities

What it does

PDF Processing

Extract text, tables, and structure from PDFs. Tier 1 uses local JS libraries for fast extraction. Tier 2 uses the Docling deep parsing engine for complex layouts, scanned documents, and OCR.

Word & PowerPoint

Read and write DOCX and PPTX files locally. Extract text, headings, tables, and slide content. Generate reports and presentations from agent output.

Spreadsheets

Parse XLSX files into structured data. Agents can query specific sheets, ranges, and named tables — no need to paste raw CSV into the context window.

Three Processing Tiers

Tier 1: local JS libraries (fast, lightweight). Tier 2: Docling deep parsing engine (complex layouts, OCR). Tier 3: remote Document Server (when you need it). You choose the tier.

File Drop Routing

Drop any supported file into a chat. The plugin detects the type and routes it to the right processor automatically. Results are injected as structured context.

How it works

From install to first use.

1
Drop or reference a fileDrag a document into the chat, or let an agent reference it by path. The plugin detects the file type and selects the appropriate processor.
2
Local processing firstTier 1 processes the file using local JavaScript libraries — no network calls. For complex PDFs with scanned pages or unusual layouts, Tier 2 (Docling) handles deep parsing.
3
Structured outputThe agent receives structured data — headings, paragraphs, tables, metadata — not a raw text dump. This keeps context windows lean and responses accurate.
4
Generate documentsAgents can write DOCX, PPTX, and other formats. Generate a report, a slide deck, or a specification document — output as a downloadable artifact.
Why local matters

Your documents stay on your machine.

Tier 1 and Tier 2 process everything locally. No file contents are uploaded to any cloud service. You control when and whether to use the remote Tier 3.

LocalTier 1 & 2 run entirely on your machine
4document formats supported natively
0files uploaded without your explicit choice
~10×fewer tokens than raw file dumps in context

Documents ships with the Studio. No extra install, no extra cost.

Get early access All extensions