Free plan: unlimited topics, generated segments each session, extend 1 extra segment per topic. Pro plan: $16.9/month (originally $29), includes unlimited segment extensions, unlimited deep-dive questions, and realtime AI voice playback. Month-to-month with no lock-in; cancel anytime.
Not given
Key features
Generate infinite audio lessons on any topic
Structured audio course with overview and deep-dive segments
Real-time playback with segment generation during listening
Web and mobile access with no download required
Sentence-by-sentence highlighting for hands-free learning
Browse hundreds of curated topics or enter custom topics
Growing knowledge trail showing mastered topics and next steps
Unlimited deep-dive questions with Pro plan
Chat to ask about work with answers sourced from your messages, calls and notes
Dictation in any text box on Mac, including browsers and code editors
Voice to talk to Jelly from any app and get spoken answers
Meeting prep, live assist and notetaker with no bot joining the call
Open Loops to track what you promised people and tick off when done
Create emails, documents, sheets and slides from conversations
Agents to do tasks on the web with transparency and payment confirmation
MCP to share chosen memory with Claude, ChatGPT or Cursor
Encrypted memory on Mac with option to use your own API key
Works offline for local processing with no screen recording
What makes it different
Never-ending: Segments keep generating as long as you listen, unlike fixed-episode podcasts
On-demand personalization: Generate content for any topic instantly, not what a host chose
Continuous learning: Knowledge trail builds across all listening sessions
Fits existing time: Learn during dead time—commutes, workouts and chores without needing dedicated study
Free to start with no payment barrier to try learning
All AI features in one subscription instead of paying for separate apps
Memory stays encrypted on your Mac, never sent to cloud unless you choose
One memory connects all features so each knows what you're working on
Scored 87.6% on BEAM public test for long-term memory accuracy
Agents work in their own browser and stop to ask before paying