Pick a mode
Owners set optimizer mode in Usage. Off leaves prompts untouched; on modes trim before each send.
Product · Optimize & Create
Context Optimizer for fewer tokens per send.
Shrink chat history and file context before each send so answers stay sharp while provider bills stay quieter.
The thread hit two hundred messages and the bill noticed before the answer quality did.
- Modes from off to aggressive
- History compaction in the background
- File snippets trimmed at read time
- Savings shown on usage
38% fewer input tokenstypical context spend after a week of trimming, per usage reports
How it works
Keep chatting
Compaction runs in the background. Your stored messages stay intact; only the outbound prompt shrinks.
Review savings
Usage reports token reductions so you can see what the optimizer actually saved.
01
Same answer. Smaller prompt.
Long threads and fat files burn tokens before the model writes a word. The optimizer budgets history and file snippets so the model still sees what matters.
- Per-workspace mode control
- Background summary chunks for aged history
- Read-time file compaction that never rewrites stored text
02
Built for BYOK teams.
When you pay providers directly, every wasted input token is your money. Optimizer settings stay owner-controlled so members never fight the dial.
- Owner-only settings
- Fail-open if settings cannot load
- Pairs with usage and spend guardrails
Questions, answered
No. Stored messages stay intact. Only the outbound prompt is compacted before each send.
Owners set it per workspace, from off to aggressive. Members never see or fight the dial.
Usage reports show token reductions per period, so the savings are a number, not a promise.