AI Automation

A local AI system (Cursor, Claude, Perplexity Computer, and similar tools) can talk to Diode Collab on the same machine: send and receive chat, and work on files that live in a Zone folder.

Chat — connect locally to the API

Enable the Remote API on this Collab instance (/sysconfig remote-api enable, then generate or view a token). Point the AI at the local JSON-RPC listener — not *.diode.link — as described in Connecting locally to the Diode Collab API.

Once that local HTTP or WebSocket URL is up and the Bearer token (or WebSocket authenticate) is in place, the AI uses the same methods as any other integration. See Remote JSON-RPC API for payloads.

Typical chat loop:

  • Send: send_message with zone_id, message_text, and either channel_id or user_name
  • Receive: WebSocket subscribe_channel / subscribe_all_channels_in_zone for pushes, or get_messages to query

Give the AI the token, the local URL, and the zone/channel ids it should use. Keep the Collab app running on that host.

Files — equip the AI with Zone folders

Chat methods do not replace the files on disk. For file automation, give the AI the Zone folder itself — the directory Collab syncs for that Zone. How to find or change that path is in Zone folder.

Point the AI’s workspace or project at that folder (or a Drive inside it). It can then read, edit, and create files the same way it would on any local project. Sharing and access still follow Secure File Sharing.

For those file modifications to surface to the rest of the team, the Zone (or Drive) must be in Upload Only or 2-Way sync — not Off/manual. See File Syncing and File Sync Policy.