Skip to content

Attachments — images and files

You can share files with your agent in a few ways. What it does with them depends on what kind of file it is.

Your agent has vision capability, so it can actually see images. This works for:

  • Screenshots — “What does this error say?” / “How should I respond to this message?”
  • Photos — “What’s in this photo?”
  • Diagrams and whiteboards — “Turn this into a structured summary”
  • UI mockups and designs — “Give me feedback on this landing page”
  • Charts and graphs — “What does this data show?”

Dashboard: paste from clipboard (Ctrl/Cmd+V) or click the paperclip icon. Max size: 10 MB per image.

Telegram: send it like any photo in Telegram.

Discord: upload as an attachment.

Slack: drag-and-drop or use the ”+” icon.

The image is sent to a vision model (Gemini 2.5 Flash) which extracts a detailed description. Your agent then works from that description. It can read text in images, identify objects, describe layouts, interpret charts.

It cannot:

  • Run OCR on handwritten notes perfectly (prints better than cursive)
  • Identify specific people by face (privacy-safe by design)
  • See every fine detail in a large image (it compresses before analysis)

Supported file types:

  • Plain text (.txt, .md, .json, .yaml, .csv, .tsv, .log) — read inline, your agent can reference the content directly
  • Office documents (.docx, .xlsx, .pptx) — extracted to text before the agent reads them
  • PDFs — extracted to text; if the PDF is an image scan, the vision model handles it
  • Code (.py, .js, .ts, .go, .rs, etc.) — read as code with syntax awareness

When you attach a file, it goes to your agent’s workspace — a persistent file storage area in your private container. The agent can reference, re-read, edit, or reuse the file in future conversations.

You can see everything in your workspace from the dashboard’s Workspace tab. Workspace storage is:

  • Trial: 10 GB
  • Standard: 10 GB
  • Pro: 25 GB

Not supported yet. If you attach an audio or video file, your agent will tell you it can’t process it directly. Voice transcription is on the roadmap.

Workaround for audio: use a transcription service first (Otter, Fireflies, Apple Voice Memos transcription), then paste the transcript.

Once a file is in your workspace, you can refer to it in future conversations by name or description:

“Look at the Q2 sales CSV I uploaded yesterday and tell me which product had the biggest drop”

“Use the brand guidelines PDF from last week to make sure this LinkedIn post matches our voice”

Your agent will find the file, read it, and use it as context. It doesn’t need you to re-attach anything.

From the dashboard’s Workspace tab, you can delete individual files or folders. Deletions are immediate and irreversible. If you want your agent to stop considering a file but not delete it, you can move it to a subfolder like archive/ and tell the agent to ignore it.