TokenWarden

opencode cost and context control

Cut token waste. Keep the AI sharp.

TokenWarden trims wasted context in opencode, keeps your AI focused, and reports observed optimization savings per session.

Start freeConnect account
$ npm install -g tokenwarden

$ tokenwarden status
Optimization: active
Privacy: local-first
Account: free seat | credited 81.7k/100k observed saved tokens

$ /tokenwarden-report
TokenWarden observed optimized-context savings
Scope: recorded optimization events only
150k observed -> 67.3k used, saved 82.7k (saved 55.1%)

Built for programmers who want proof, not vague promises.

Measure observed waste

See would-have-used tokens, actually used tokens, saved tokens, and percentage saved for recorded optimization events.

Control context

Optimizes noisy context so your AI gets what matters without dragging in the whole repo.

Stay local

Your coding context stays on the developer machine by default.

Save up to 90% on optimized context events.

Free includes

100,000optimized tokens per month

for individuals to test and explore.

Estimate observed context savings

Enter estimated monthly optimizable usage. This is not a full provider bill forecast.

Savings are estimated at 90% for observed optimization events only. Normal prompts and provider-billed tokens outside those events are not included.

Observed-event estimate$135.00per month, before provider billing variance

Pricing

Choose monthly or save with yearly billing.

Save 20% with yearly

Free

For individuals getting started.

$0/month

Includes:

  • 100,000 optimized tokens per month
  • Basic optimization features
  • Community support
  • Single machine activation

Standard

For professionals and teams.

Most popular
$96/seat/year

Everything in Free, plus:

  • Uncapped optimized tokens
  • Advanced optimization features
  • Priority support
  • Multiple machine activations
  • Team management and analytics
Measured savings

Track optimized-context savings without claiming every provider-billed token.

Better context

More room for what matters. Less noise.

Built for privacy

All optimization happens locally. Your code stays yours.

Reliable enough

Local-first reporting that keeps savings visible across sessions.

Frequently asked questions

What are optimized tokens?

Tokens TokenWarden avoided in recorded optimization events, such as context it actually changed or optimized.

How is savings calculated?

TokenWarden compares would-have-used tokens with actually used tokens for recorded optimization events. It does not count every model input, model output, cache token, or provider-specific billing unit.

Can I change plans later?

Yes. Start monthly, move yearly for 20% savings, and adjust paid seats from your account page.

All plans include local-first privacy. No source code is used to train models. Ever.