On the prices each vendor publishes about itself, Claude Code starts at Not read and Onlook at Not read. Both take your own API key, so on either one the subscription and the model bill are separate decisions. Onlook is open source under Apache-2.0; the other is proprietary. Every figure on this page was read from the two vendors' own pages in last night's run rather than written down once, which is the only reason it is still the current answer. ToolDrift sells neither and takes no referral fee on either.
| COMPARED ON | ||
|---|---|---|
| ASSERTED · WHAT EACH VENDOR PUBLISHES ABOUT ITSELF | ||
| Category | CLI agent | Design to code |
| Vendor | Anthropic | Onlook |
| Status | Active | Active |
| How it charges | Subscription | Hybrid |
| Entry price | Not read | Not read |
| Default model | anthropic/claude-opus-4.8 | None, you choose |
| Your own key | Yes | Yes |
| Source | Proprietary | Open source · Apache-2.0 |
| Plans | No named tier on Anthropic's pricing page could be read with a price beside it. | No named tier on Onlook's pricing page could be read with a price beside it. |
| MEASURED · COUNTED ON THE LAST NIGHTLY RUN | ||
| Stars | 147.9K | 26.8K |
| Maintenance | 94/100 | 92/100 |
| Days since last push | 0 | 30 |
| Top model in it | Anthropic: Claude Opus 4.8 · $10.00 | not measured |
| Its volume | 338B tokens | not measured |
Is Claude Code or Onlook cheaper?
On the cheapest option each vendor publishes: Claude Code is Not read and Onlook is Not read. Treat both as an allowance rather than a bill, every tool in this category meters model usage on top of the plan, and which one costs more in practice depends on the models you drive and how hard. The tier prices are the part that can be checked from outside, so they are the part this page prints.
Can I use my own API key with Claude Code or Onlook?
Claude Code: yes. Onlook: yes. It is the row that decides whether the subscription price is most of what you pay or a fraction of it: a tool that takes your own key charges for the software, and one that does not is reselling inference with a margin you cannot see.
Are Claude Code and Onlook still maintained?
Active and Active as of last night's read. Claude Code's repository was last pushed 0 days ago and it scores 94/100 on maintenance; Onlook's repository was last pushed 30 days ago and it scores 92/100 on maintenance. This is the question a written comparison cannot answer, because it was true on the day it was published and nobody went back.