Skip to main content
Spaceduck’s agent loop can call tools automatically during a conversation. When the assistant needs external information or wants to take an action, it invokes a tool, reads the result, and continues the conversation.

Available tools

How tools work

The assistant decides when to use a tool based on the conversation. You don’t need to invoke tools manually — the agent loop handles it:
  1. You ask a question or make a request
  2. The assistant determines a tool would help
  3. The tool runs and returns a text result
  4. The assistant reads the result and responds
Tools can chain — the assistant might search the web, then fetch a specific page, then summarize the content. This happens automatically across multiple rounds.

Tool results are text

All tool results are plain strings that the LLM can read — including errors. There are no structured schemas or silent failures. If a tool fails, the assistant sees the error message and can explain what went wrong.

Configuration

Tools are configured in Settings > Tools. Some tools require external dependencies to be installed before they appear.
ToolRequiresAuto-registered
BrowserPlaywright + ChromiumYes (if installed)
Web SearchBrave API key or SearXNG URLYes (if configured)
Web FetchNothingAlways available
Document Scanningmarker_single on PATHYes (if installed)
The assistant also has built-in tools for reading and updating configuration (config_get, config_set) — you can ask it to change settings during a conversation.