Same intelligence.
A different conversation.
MUSE on Telegram uses the shared server AI backend. Configuration and delivery are not verified until the tests below pass.
Integration requires deploymentBot commands
/startMeet MUSE and get started/helpCommands, privacy, and search/clearDelete this Telegram conversation/search questionSearch with Tavily, then answerDeployment reality
Functions can make short server-side API requests. Persistent workers and a public Telegram webhook endpoint are not established by the supplied capabilities. The included external relay must run on a host you control, with a verified function invocation URL.
The web app works without Telegram. The external relay works without the web UI. See DEPLOYMENT.md for required setup and privacy constraints.
Secure configuration
Only an app owner can inspect secret names or send a test message. Values are never displayed.
Live verification
These buttons perform actual calls and consume quota. PASS is shown only after a real response. History checks browser storage; private cloud sync is a separate, untested requirement.
AI response
NOT TESTED
Web search
NOT TESTED
Chat history
NOT TESTED
Telegram delivery
NOT TESTED
Free-tier boundaries
Built-in AI: 100 requests per app per hour, up to 8,192 output tokens, with a 60-second timeout. Built-in live search needs Pro and is not used here. Tavily uses a separate server-side key and account quota. Shared OpenAI responses use your provider billing. Functions have a 30-second limit; the relay needs separate persistent hosting.