Billing & usage
Plans, usage counting, and what's free.
Short version: the terminal is free. The hosted AI agent runs on credits — every plan (including Free) includes a monthly credit allotment, and you can always bring your own provider key for unlimited, 0-credit use. No waitlist.
What's free
Everything in the terminal is free to download and use on macOS, Windows, and Linux: blocks, the command palette, splits and tabs, chips, the built-in code review and editor, CLI agent sessions, SSH, and global search. None of it requires an account, and none of it ever costs credits.
Agent access needs a free Mel account — sign-up is self-serve and instant. On the Free plan you get 150 hosted-AI credits a month on the hosted models with no API keys of your own (provider keys are held server-side on the Mel relay); bring your own provider key for unlimited, 0-credit use, or upgrade for a bigger monthly allotment. Create an account from the account sign-up flow.
Credits
1 credit = $0.01 of the underlying provider's list-price API cost. A turn is billed for the tokens it actually uses on the model that answers it, rounded up, with a 1-credit minimum — so a cheap value-tier turn costs about 1 credit and a frontier-model turn costs more. The automatic router keeps everyday turns on cost-efficient models, so credits go a long way.
Monthly allotments (reset each month):
| Plan | Price | Credits / month |
|---|---|---|
| Free | $0 | 150 |
| Pro | $20 / mo ($18 annually) | 1,500 |
| Max | $200 / mo ($180 annually) | 18,000 |
| Business | $50 / seat / mo ($45 annually) | 1,500 per seat (org-pooled) |
- Bring your own key = 0 credits. A turn on your own provider key never touches your allotment.
- Shell commands and CLI coding agents are always free — they don't use Mel credits.
- Out of credits? The agent pauses with a clear message; your shell and your own key keep working. Buy an add-on pack, or wait for the monthly reset.
Add-on ("Personal") credit packs
On a paid plan you can top up any time with prepaid packs — 400 ($10) · 1,000 ($20) · 3,000 ($50) · 6,500 ($100). Add-on credits roll over and expire one year after purchase; they're used after your monthly base credits are spent. Optional auto-reload buys a pack automatically when you run low (respecting a spend limit you set).
The Billing and usage settings page
Open it from the avatar menu (Billing and usage) or via Settings. It shows, live from your account:
- Your plan and current credit balance.
- Two balance cards: Plan credits (base allotment remaining + the monthly reset date) and Personal credits (add-on balance + expiry).
- Upgrade options for higher tiers, and buy add-on credit packs (paid plans).
- Your monthly spend limit and how much you've purchased this month, plus the auto-reload toggle.
Seeing your usage
- Every agent response ends with an expandable line —
<model> · N tokens ›— that opens a usage summary: the model used (and why the auto router picked it), tokens in and out, tool calls, context percentage, time to first token, and total time. - When your base balance runs low, the agent shows a small "low on credits" banner with a rough estimate of how many turns are left.
Per-plan feature caps
Beyond credits, a few server-cost features scale with your plan (bring your own key or an admin account bypasses them):
| Feature | Free | Pro | Max | Business |
|---|---|---|---|---|
| Cross-session memory recall | 30 days | 1 year | Unlimited | 1 year |
| Images per message (vision) | 3 | 10 | 10 | 10 |
| Parallel orchestration agents | 1 | 3 | 8 | 3 / seat |