Academy · Token cost
How do AI tokens work?
What are AI tokens and how do I keep costs down?
Beginner
A token is roughly three quarters of a word. Every back-and-forth with Penny consumes tokens, which cost money on the model providers' end (whether you pay them directly via your own key or pay Pendraic for managed access). Three things keep costs low: (1) Penny answers a lot of common questions from canon for free (the Academy you're reading is the visible mirror of that brain), (2) shorter conversations cost less than long ones, and (3) using the inline editor for prose generation is cheaper than asking Penny to draft prose in chat.
Standard
The Pendraic chat layer caches your project context across turns, which keeps the per-turn cost low even on long agentic builds. The biggest cost spikes come from huge audits (full manuscript pass) and image generation. Both are gated; Penny will tell you the rough cost before kicking off a heavy job.
What to try
- Open AI Settings to set up keys
- Check your wallet balance
Related questions
- Should I use BYOK or Managed AI?
- What's the cheapest model?