context
spa
Sandbox
API Docs
Try the sandbox
Sign in to get a free API key and credits. No card required.
Sign in with Google
Load example…
Meeting transcript
Customer feedback
Agent conversation
Research notes
↻
Settings
Compress
Target
32K
64K
128K (max)
Max output tokens. Smaller = more aggressive compression.
Coherence pass
Restore grammar via LLM (+10–20s). Off by default for speed.
Preserve code
Keep code blocks verbatim.
Evaluate
Run quality scoring on result (+~3s).
Skill
Natural language instructions for the compressor.
Input
Compressed
Copy
Press
Compress
or
Ctrl+Enter
—
reduction
—
tokens in
—
tokens out
—
coherence
—
runtime