What you can ask
Once an assistant is connected, talk to it the way you would talk to a colleague at the desk. It reads your station and prepares the work.
- Build a rotation from my soul and house tracks.
- Fill the gaps in next week’s schedule and leave them as drafts.
- Why did listeners drop on Friday night?
- Which tracks are missing cover art or tags?
Connect in two steps
In the Desk, open AI and agents and choose Connect an assistant. Pick Claude, ChatGPT or another app, choose which stations it can reach and what it can do, then add the connection in your assistant.
Each connection ends automatically after 30 days, 90 days or a year, and you can disconnect it at any time.
You choose what it can do
Every connection has one permission level. Pick the smallest one that does the job.
- View only: reads audio, playlists, the schedule and listener numbers, and changes nothing.
- Playlists and schedule: uploads audio, builds playlists and drafts programmes, and can start and skip automatic playback.
- Station manager: everything above, plus presenters, alerts and the directory listing.
- Media delivery: for services such as Syndifeed. Sends shows to your station and nothing else.
You stay in charge
An assistant prepares programmes as drafts, and you decide what goes on air. Billing, team members and removing a station are never available to an assistant, whatever its permission.
SGNL checks every request on the server, so an assistant can’t reach further than its connection allows or see another workspace.
For developers
Create a developer token in AI and agents to use the same server from scripts and other MCP clients with a bearer header. The Desk shows the server address when you create it.
On browsers that support WebMCP, the SGNL Desk home page also offers visiting agents a few public tools: comparing plans, recommending one, explaining live broadcasting, checking whether a move fits, and opening the beta form. They use only published information and can’t buy anything. A short product summary for agents is at sgnldesk.co/llms.txt.