How credits work
The unit
Section titled “The unit”1 credit = $0.001 of compute. So 1,000 credits = $1.
Your Pro plan includes a monthly credit allowance that refreshes each billing cycle (it does not roll over). Credit packs you buy separately are one-time top-ups that are used when the monthly allowance runs out.
What spends credits
Section titled “What spends credits”- Messages and agent runs, priced by the model and how much it had to think
- Web search and page reading
- Advanced tools: browser, computer use, mobile emulator, SMS and calls
- Image, video and voice generation
- Scheduled tasks — every run costs like a normal run
- Subagents, which are simply more agents working
Roughly: an advanced model costs about 50 credits per 1,000 tokens of input plus output, so 1,000 credits is around 20,000 tokens — several medium chats, or one longer agent task. Cheaper models use noticeably fewer credits.
What doesn’t spend them
Section titled “What doesn’t spend them”- Work done through a connected Claude, Cursor or Codex subscription — that’s billed by them, not by TeleClaw. See Connect coding agents.
- On the main @claw bot, simple answers in group chats, up to a fair-use allowance of 250 credits per 6 hours per person. Image, video and other generation still costs credits.
Watching your balance
Section titled “Watching your balance”The Mini App shows your credit balance and a usage screen with what was consumed. Free accounts also get a one-time welcome grant, and reward tasks in the app can earn more.
Keeping the bill down
Section titled “Keeping the bill down”- Prefer a cheaper model for routine chatter; save the advanced one for real work.
- Check the frequency of scheduled tasks — a job every minute is 1,440 runs a day.
- Turn off tools an agent doesn’t need in its settings.
- In groups, leave I pay for messages off unless you intend to sponsor everyone.
Running low? Buy a credit pack or move to a higher tier — see Manage your subscription.