Agent

Reading & Summarizing Files

Use Poly's AI agent to read, summarize, and answer questions about your files.

The Poly agent can read the contents of any file in your storage and respond to questions about it. You don't need to open the file, copy text, or paste it into a chat — just tell the agent what you want to know.

Asking about a specific file

The most direct way is to select a file in the browser and then open the agent:

  1. Click a file to select it.
  2. Press ⌘ / (Ctrl+/ on Windows and Linux) to open the agent.
  3. The agent is now aware of that file and ready to answer questions about it.

Try asking:

  • "Summarize this document in three bullet points"
  • "What are the key dates mentioned in this contract?"
  • "What is the main argument of this paper?"
  • "List all the line items in this invoice"
  • "What's the resolution and camera settings for this photo?"

Asking about files without selecting them first

You can also reference files by name or description in the agent chat:

Summarize the Q3 financial report
What does the terms of service document say about data retention?
Give me the key findings from the user research report uploaded last week

The agent searches your storage to find the relevant file and reads it as part of answering your question.

Asking about multiple files at once

The agent can read and compare several files in one turn:

Compare the proposals from Vendor A and Vendor B — what are the main differences?
Give me a summary of all the meeting notes from this quarter
What themes come up across the feedback documents in the Research folder?

For large collections, the agent may process files in batches and synthesize the results.

Supported file types

The agent can read:

  • DocumentsPDFs, Word documents, presentations, spreadsheets, and ebooks, page by page
  • Text & code — any text-based format, verbatim
  • Images — the agent sees the image itself: it can describe content, read visible text, and answer questions about photos
  • Audio — through the automatic transcript of the recording
  • Video — through the automatic transcript of its audio track
  • Links — the content of the linked page, fetched and read like a document
  • YouTube videos: bookmarked YouTube links, through the video's metadata and content

Audio and video become readable as soon as transcription completes after upload — the details panel shows each file's processing status. The full per-format story is in Supported File Types.

What the agent does not do

The agent reads your files; it does not share them with anyone else or use them to train Poly's models. See Privacy & Security for details on how your data is handled.

© Poly Corp. 2026