> ## Documentation Index
> Fetch the complete documentation index at: https://docs.cohesive.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Your first session

> Turn one source file into a reviewed result on your first Canvas.

Create a Canvas, upload one file, and ask the agent to turn it into a useful summary.

Before you begin, sign in to [Cohesive](https://app.cohesive.ai) and have a source file ready, such as a project brief, meeting transcript, or proposal. Each upload can be up to 100 MB.

<Steps>
  <Step title="Create a Canvas">
    Open **Workspaces** in the left rail. Right-click **My Workspace** and choose **New canvas**, or use the rail's create menu. Give the Canvas a name that describes the work.
  </Step>

  <Step title="Add a source">
    Drag your file onto the Canvas. Its card appears while the file is processed. Wait for processing before asking questions that depend on the full contents.

    You can also press `N` to add a note with the context you want the agent to use.
  </Step>

  <Step title="Select the source and open Quick Chat">
    Click the source card, then press `C`. The selected card is included as context.

    If you want to use an existing subscription, [connect ChatGPT or Claude](/product/account/model-providers) and choose that agent before sending the first message. You can also continue with Cohesive.
  </Step>

  <Step title="Ask for a specific result">
    Try: *"Summarize this document in a note: key decisions, open questions, and next actions. Include references to the source for important claims."*

    For a longer output, ask for an editable document instead of a note.
  </Step>

  <Step title="Review and revise">
    Open the result and check its claims against the source. Continue in the same chat: *"Separate confirmed decisions from proposals, and add an owner to each action where the source names one."*
  </Step>

  <Step title="Organize and keep the work">
    Select the source and result, then choose **Wrap in section** from the right-click menu. Name the section.

    Use **Add to workspace** on a file you want to reuse across Canvases. [Share the Canvas](/product/collaboration/sharing-canvases) for live work with teammates, or [publish the document](/product/files/publishing) as a public snapshot.
  </Step>
</Steps>

## If the result misses the point

Check that the right cards are selected, the file has finished processing, and the prompt names the output you want. Ask the agent to identify missing information rather than fill it in.

## Continue with a real task

<Columns cols={3}>
  <Card title="Document briefing" icon="file-lines" href="/product/guides/document-briefing" />

  <Card title="Meeting preparation" icon="calendar" href="/product/guides/meeting-prep" />

  <Card title="Parallel research" icon="magnifying-glass" href="/product/guides/research-in-parallel" />
</Columns>