Every condense plan comes with a weekly pool of tokens saved. Before Condense Lite, new compaction stopped once that pool was empty. Now condense keeps managing your context, saving up to 20% of each prompt until the pool refills.
We announced Lite on August 28 and turned it on in production on September 10.
Lite isn't a plan you switch to. Every plan lands on it when its pool empties, and goes back to full strength when the pool refills at the start of the next week.
You lose some of the savings, never the service. Short sessions come out the same either way, because they never reach Lite's limit. The gap opens up the longer a session runs. At full strength, savings grow with your sessions, up to about 68%, and a bigger pool keeps you there for longer.
Lite works with Claude Code, Codex and OpenCode. Your dashboard shows when it's running and when your pool refills.
What free means
Lite costs nothing extra, on any plan. It's unmetered and never runs out. A few things still apply.
- Account
- You need one. The Free plan is $0 with no card required, and includes 25M tokens saved a week at full strength, then Lite.
- Provider bill
- condense sits in front of your provider. You still pay Anthropic or OpenAI directly, just for fewer tokens.
- Rate limits
- Your plan's per-minute limit stays in place, from 20 requests a minute on Free.
- Direct API
- Lite comes with plans. Calling the API directly is priced separately.