using-arcade-tools

Using Arcade tools

Quick start

Arcade_SelectTools(tasks: ["<one plain-language task>"])   # candidates + input_schema inline
Arcade_UseTool(tool_name, inputs, query_id)                # execute; tool_name verbatim

Tools are discovered at runtime — there is no static list. Treat discovery as invisible plumbing the user shouldn't think about.

Prefer delegating

When a task needs an external service, hand it to a subagent so the discovery noise stays out of the main conversation:

• General tasks → arcade-operator
• Email → inbox-agent
• Calendar / scheduling → schedule-agent

Calling the tools directly

1. Arcade_SelectTools(tasks=[...]) — one tasks entry per logical workflow. Returns candidates with their input_schema inline.
2. Arcade_UseTool(tool_name, inputs, query_id) — use tool_name exactly as returned (Toolkit_Action form, no @version, no period); forward query_id.
3. Arcade_Apps — to see or disconnect the user's connected apps (see the managing-arcade-apps skill).

Behavior

• Don't narrate discovery ("let me search for a tool…"). Deliver the outcome.
• Don't dump schemas or present a list of tools for the user to pick from.
• A tool that needs an app the user hasn't connected returns a one-time sign-in link — present it, then retry once they confirm. Never loop.
• Prompt only when a genuinely required input is missing (which channel, which repo, which recipient).