
Send content to any screen
agentView lets your AI update any browser-capable screen (a TV, tablet, kiosk, or signage panel) by talking. Ask your AI to put a daily lunch menu on your bistro display, a sales dashboard on the office wall, or a waiting-room board on your practice screen, and the layout appears live within seconds.
Example path
https://agentview.de/mcpAgentView MCP: digital signage — manage displays, send HTML/URLs, install store templates, data slots, assets, organizations.
STEP 0 — AUTH FIRST: protected tools return error 'auth_required' without a login. If you hold no session_request_id from THIS conversation: 1) (scope 'display_setup' if no screen yet) → 2) show the loginUrl as a normal clickable link in the user's language (never a code block) → 3) poll every ~2s until status 'active' → 4) pass session_request_id on EVERY later call; it survives dropped OAuth headers. No auth needed: , search, fetch, , , , , (template mode), , , , authenticate.
AUTH RECOVERY: error 'invalid_token' (token dropped mid-conversation) → silently retry with the session_request_id from the latest ; never show raw auth errors. No session_request_id at hand → restart STEP 0 in plain language.
No screen yet: user opens display.agentview.de on the device that should show it and reads you the 6-char code → (name optional). Chat device itself as screen: , give managedUrl.
See or show a display: ALWAYS (inline widget).
Send content: → (or ); slot 'idle' sets the fallback.
After first content: offer ONE way to keep it alive (whole week in a template, or self-refreshing live data). Never end on a dashboard link.
None in the store listings or on the server

