What you pay
$9.99
USD / month
One individual plan with local optimization, evidence restoration, and a transparent causal ledger.
Local-first Claude Code efficiency
Keep noisy logs and oversized tool results out of the model context. Preserve the exact raw evidence locally, restore it on demand, and see what your plugin prevented.
Estimated usage preserved
1.84M
A clear value comparison
What you pay
$9.99
USD / month
One individual plan with local optimization, evidence restoration, and a transparent causal ledger.
What you preserve
$100+
API-equivalent estimate / month
Based on a measured high-usage profile. Results vary by model, cache state, task mix, and session behavior.
View methodologyHow it works
The plugin intercepts oversized tool output, stores the original bytes locally, and returns a compact, auditable view to Claude Code.
Recognized logs, test runs, JSON, and repetitive terminal output are reduced before they enter the next model turn.
Every compacted result keeps a content-addressed handle, checksum, provenance, and bounded read path.
See raw bytes prevented, tokens estimated, latency added, restoration activity, and API-equivalent methodology.
Causal ledger
This example uses real product data shapes: original size, returned size, estimated tokens avoided, latency, state, and restorable handle.
| Optimization | Original | Returned | Estimated avoided | Status | Evidence |
|---|---|---|---|---|---|
| Pytest output firewall | 486 KB | 2.3 KB | 118.4K | Prevented | qg_8c1f…72a9 |
| Repository snapshot read | 64 KB | 8.1 KB | 13.7K | Restorable | qgr_4d9a…3b10 |
| Unique error payload | 5.2 KB | 5.2 KB | 0 | Passthrough | raw |
Simple pricing
The optimization path stays local by default. Cloud services are limited to account, license, and opt-in aggregate product data.
Methodology note: Savings are estimates derived from measured bytes prevented and public API-equivalent token pricing. They are not a fixed Claude subscription quota or a universal guarantee.
Individual