The Claw Compactor
Crush your OpenClaw agent context by 98%. 5 compression layers, pay-per-run with $cCLAW. The first tokenized utility in the claw ecosystem.
Origin
aeromomo built the compactor. We built the meter.
claw-compactor is an open-source Python tool by aeromomo, built for the OpenClaw ecosystem. It compresses AI agent workspaces — memory files, session transcripts, sub-agent context — using 5 rule-based compression layers. No LLM needed. Python 3.9+. MIT licensed.
$cCLAW is the fuel for the hosted compression API. Every run burns tokens proportional to workspace size. The CLI stays free forever. The token powers the managed service: one API call, instant results, no local setup.
How It Works
5 Compression Layers
Data flows through each layer sequentially. Each technique targets a different type of redundancy. Together they achieve 50%+ savings on a typical workspace.
Dedup lines, strip markdown filler, merge sections
Auto-learned codebook, $XX substitution
Session JSONL → structured summaries
Path shorthand ($WS), IP prefix, enum compaction
Ultra/medium/light abbreviation levels
Real-world savings
Combined with caching
95%
effective cost reduction
50% compression + 90% prompt cache discount
Burn Model
Every run burns $cCLAW
Bigger workspace = more tokens burned. Volume discounts kick in automatically. No tiers, no staking gates. Just use it.
Compress
1K tokens
500 $cCLAW
Single file cleanup
Compress
10K tokens
4,000 $cCLAW
20% volume discount
Compress
100K tokens
30,000 $cCLAW
40% volume discount
Compress
1M tokens
200,000 $cCLAW
60% volume discount
The open-source CLI is always free. Burns apply only to the hosted API.
FAQ