Tools

This page is for controlling what an assistant can do beyond answering from a prompt. A tool function is an operation the LLM may call while producing an answer — searching a datasource, fetching an instruction document, updating a checklist, or asking another assistant. The page has two halves: publishing this assistant as a tool for others, and managing the tools this assistant may use itself.

Tools Config

Publish this assistant

Assistants can call each other: a published assistant appears as a tool function named chat_<name> that other assistants — and external MCP clients — can invoke.

FieldDescription
DescriptionThe text callers and their LLMs see. Publishing has no effect while this is blank — write it first.
Expose as toolLets other assistants call this one. Each calling assistant must still enable the function on its own Tools page.
Expose via mcpPublishes the assistant on the MCP server for external MCP clients.
Required fieldsComma-separated parameter names a caller must supply. Blank uses the default (normally q).

The What callers receive expander at the bottom shows the exact tool description and parameter schema offered to callers.

Tool use

The Tool use section holds the master switches: Enable tool usage turns tool calling on or off for the assistant as a whole, and Max iterations bounds how many rounds of tool calls a single answer may take.

Available tools

Every tool function the assistant can see is listed here, grouped by capability — Search, External sources, Tasks, Memory, Assistant, Utility and so on. The All / Enabled / Disabled buttons filter the list by resolved state.

Available tools

Each row shows the function’s name, the description the LLM receives, and an enable dropdown:

ChoiceDescription
Default (Enabled) / Default (Disabled)Inherit the function’s own default. The parenthesis shows what inheriting currently means.
EnabledAlways on for this assistant.
DisabledAlways off for this assistant.

Expand Details on a row to see the full description and the parameters the LLM receives, and to set a description override — a replacement description that applies to this assistant only. Use {description} in the override to insert the tool’s underlying description. Leave it blank to inherit.

Everything on this page saves automatically when a field changes — a small ✓ next to the field confirms it. The settings belong to the individual assistant; the shared defaults for search functions are managed per config set on the Retrieval page.