Send attachments and context
You can attach files to any message in MultiClaw. The agent reads the file content as part of your message, so it has the context it needs to help you accurately.
File content you attach is sent to the AI model configured for your agent — which may be hosted by a third-party provider such as OpenAI or Anthropic — to generate a response. Do not attach files containing sensitive personal data unless you have reviewed your agent's model configuration and confirmed this is appropriate.
Attach a file
When chatting with a local agent (connected to your local OpenClaw installation), there are three ways to attach files:
- Paperclip button — click the attachment icon in the composer and select one or more files from the file picker.
- Drag and drop — drag files directly onto the composer area and release.
- Paste an image — paste a screenshot or copied image from your clipboard directly into the composer (images only; 10 MB limit per pasted image).
When chatting with an agent on a cloud desktop, only the paperclip button is available.
After attaching, files appear above the composer. Add your message if you haven't already, then press Enter or click Send. The agent receives your message and the file content together.
Supported file types
| Type | Extensions |
|---|---|
| Plain text | .txt, .md, .csv |
| Images | .jpg, .jpeg, .png, .gif, .webp |
| Code | .js, .ts, .tsx, .jsx, .py, .rs, .go, .java, .rb, .c, .cpp, .h, .sh |
| Config and markup | .json, .yaml, .yml, .toml, .xml, .html, .css |
PDF files are not currently supported.
Image analysis depends on the AI model configured for your agent. If your model does not support vision, the agent will not be able to describe or interpret image content. When chatting with a local agent, a warning appears in the composer if your model may lack image support.
Context window limits
Every AI model has a context window — the total amount of text it can process at once. This includes your conversation history, your current message, and any files you attach.
If the combined size of your conversation and attachments exceeds the model's context window, the agent will notify you. It cannot process the request until the context fits within the limit. This is a constraint of the underlying AI model, not MultiClaw itself.
When pasting an image from your clipboard, the maximum size is 10 MB per image. Files attached via the file picker or drag and drop do not have an enforced size limit in the app, though very large files may still exceed your model's context window.
Tips for large files
If a file is too large to attach in full, try these approaches:
- Paste a relevant excerpt instead of the whole file — copy the section the agent needs and include it directly in your message.
- Break it into parts — send the file in sections across multiple messages.
- Summarize first — ask the agent to work with a short description of the file's contents, then send specific sections as needed.
For large codebases, attaching an entire project will almost always exceed context limits. Identify the specific files or functions relevant to your question and attach only those.
Related articles
Start a conversation
Open the Conversations screen, start a new chat from the welcome greeting, pick an agent inline, and continue past sessions from the sidebar.
Quick chat vs tasks
Understand when the agent replies conversationally and when it takes action on your behalf.
Conversation history
All your past conversations are stored in the left panel — you can delete and export them at any time.