Buying snapshot
Scroll sideways to see all columns. With a keyboard, focus this table and use the arrow keys.
| Decision point | What capsul states |
|---|---|
Who publishes it | Adrien Pennetier, a sole trader trading as PENRA. |
Agents it drives | Claude and Codex, through the CLI you are already signed into, so it needs no API key to get started. |
What leaves your machine | Your task word for word, the files you point at, short extracts of the repository and whatever the agent then asks to open; every request shows up in your terminal. Whole files nobody asked for, past sessions, keys and environment variables are never sent. |
What capsul's server sees | capsul says its server is told that a prompt ran, never the prompt, the repository or the reply. The savings ledger is a file on your own disk. |
Languages | JavaScript and TypeScript are the most precise; Python, Go, Rust, Java and Ruby are well covered; any other text file still works. |
Token budget | capsul stops at the budget ceiling and lists what it left out; raise it with --budget or name a file with --open. |
Our take
Try capsul if you already pay for Claude or Codex and want smaller requests. The free ten prompts on your own repository let you measure the saving on your own tasks before you pay.
Evaluate capsul
Comparisons
Where it stands out
A boundary you can read. capsul lists what it sends and what it never sends, so you can check the whole boundary before a request goes out.
No second bill. It drives the Claude or Codex CLI you are already signed into, so adopting it adds no model bill of its own and no API key to manage.
Trade-offs to weigh
The headline is an input figure from the vendor's own test. The 67% median is the reduction in cache-weighted input on five short questions, and capsul says savings vary with the task and the cache. Treat it as a starting point, not a forecast of your bill.
Caching is a stated goal, not a measured result. capsul says what it sends is built so a prompt cache stays useful, and that caching as a lever of its own is on the roadmap and will be measured before it is claimed.
The Pro price is a launch offer. The terms let the publisher change prices at any time, so check the price at checkout.
Pricing and plan conditions
Free gives ten trial prompts with no card. Pro is €5.99 per month as a launch offer (original price €9.99) with unlimited prompts. Team starts at €20.99 per month with 15 seats included and €3.50 per additional seat up to 30. No VAT is charged, and the publisher may change prices at any time, with rates in force during a subscription staying applicable until it expires. A subscription can be cancelled from the account and stops at the end of the paid period.
Public signals
About these signals
Verified factual signals from public sources. They indicate observable activity or interest, not total adoption, product quality, or cost.
29 npm weekly downloads0 vulnerabilities across 1 package
See all signals from 2 sources
npm · @penra/capsul@1.4.8
Alternatives to capsul
Where capsul sits against the products teams weigh it up with.
Related technologies
Normally used together rather than chosen between, so these are not alternatives.
- CCDash
- Both target Claude usage limits from different sides. CCDash reads Claude Code's local session files and tracks the 5-hour and 7-day quotas, burn rate and rate-limit prediction; capsul lowers what each request sends. Watching quota does not reduce a request, and sending less does not show how close the next throttle is, so a developer hitting limits can use both.Applies to: Hitting Claude subscription limits: watch quota and burn rate with CCDash, lower per-request context with capsul.
- Claude Usage Tracker
- Both address the cost of Claude Code and Codex, from opposite ends. Claude Usage Tracker reads local session files and reports what was spent across tools; capsul sends less context per request and keeps its own local ledger. A developer can run both: one shows where tokens went, the other lowers what each request sends. Neither replaces the other.