Token Optimizer
 

Saved so far

Loading…
net context tokens avoided at the MCP transport
Loading… direct API-price equivalent saved
Loading… smaller than it would have been

What it has learned

Explore what it knows →

Token accounting

Gross payload reductions, later expansions, observed returns, quarantined reports, and graph estimates stay separate.

provenance-gated on this machine

Loading measured token accounting…

Provider usage & price

Native uncached input, cache reads, cache writes, and output from local CLI transcripts, deduplicated by provider request.

Loading

Loading native provider usage…

CLI & model Requests Uncached input Cache read Cache write Output API-price equivalent

OrcaRouter

One OpenAI-compatible endpoint in front of many providers. The key belongs to your OrcaRouter account and is billed to it.

Loading

OrcaRouter – API

Paste a key you already hold. Nothing opens.

Starts with sk-orca-. It is sent to this local server and stored under your token-optimizer home, never in the page.

OrcaRouter – Auth

Sign in with your OrcaRouter account. No key to copy.

Uses OAuth 2.0 with PKCE. A browser tab asks for consent and the key comes back to this machine; no client secret is involved.

Model for optimizer calls

Loaded from this workspace's own model list. No free-text entry: a name that is not in the list is not offered.

Checking the catalog…

Measured by agent

New optimizer measurements use MCP handshake identity. Historical rows without identity remain unattributed instead of being guessed.

Loading agent attribution…

Observed activity by source

Cross-client observations are used instead of a single CLI's stale session file.

Which actions ran most

Ranked by observed calls; no token amount is invented when a CLI omits it.

Recent measured optimizer activity

Verified reductions, observed returns, and excluded historical claims, most recent first.

Every action, counted

Action Times used Context used Cost equivalent Verified saved Excluded report