Claude Code plugin · private beta

Spend Claude where it counts.

Havrin moves the routine steps of a Claude Code session to cheaper models and your own machine. Claude keeps the hard parts. Your tests check every result.

Join the waitlist How it works

01 · How it works

Every step finds its own path.

Route

Searches and file exploration that Claude hands to a subagent run on Sonnet instead of Opus. Built.

Trim

Long tool output is cut before it enters the conversation, because Claude rereads all of it every turn. Testing.

Verify

On a machine with a capable GPU, a local model drafts routine fixes and your tests decide what is kept. Research.

02 · Where usage goes

Most of the bill is Claude rereading.

58%

of cost was rereading cached context, which happens on every turn.

48%

of cost came from subagents, all of them on Opus. Havrin moves these first.

225k

tokens of context on an average request, across 10,915 requests.

30 days of one developer's sessions, weighted by cost

03 · Free meter

See your own number first.

The meter replays your last 30 days from local Claude Code logs and estimates what each change would have saved. Free, open source, and nothing leaves your machine.

havrin meterexample output

$ pipx install havrin
$ havrin meter --days 30 --what-if

cost by model    opus 100%  sonnet 0%
subagents        48% of cost

what-if
routine subagents on Sonnet   up to 19%
context capped at 150k        up to 19%
both                          up to 38%

View on PyPIPython 3.10+, no dependencies

04 · Waitlist

Get the beta.

One email when it opens. Every change ships only if it matches plain Claude Code's pass rate, and the benchmark is published first.

Your email is used only for beta news and never shared.