tokenqrusher
oleh qsmtco
Coding Agents & IDEsRingkasan
Token optimization system for OpenClaw reducing costs 50-80%.
Deskripsi
Apa yang Dilakukan Skill Ini
tokenQrusher reduces OpenClaw API token costs by 50-80% through two hooks: one that loads only the workspace files needed for a given message's complexity, and one that stretches heartbeat check intervals. All processing runs locally with no external calls.
Rather than manually trimming prompts, it classifies each message automatically and strips irrelevant context before the agent ever sees it.
Kapan Menggunakannya
- Cutting monthly API bills for a busy personal OpenClaw agent
- Loading minimal context for quick greetings and one-liners
- Suppressing heartbeat polls during overnight quiet hours
- Checking which files would load for a specific prompt before running it
- Reducing API call volume for calendar and email polling
Contoh Alur Kerja
User asks: 'hi'
- 1
token-context hook fires on agent:bootstrap
- 2
Classifies message complexity as 'simple' with 95% confidence
- 3
Filters workspace files down to SOUL.md and IDENTITY.md
- 4
Passes filtered file list to the agent instead of full context
Agent bootstraps with 2 files instead of 7, saving up to 99% of context tokens for that message
Instalasi
clawhub install qsmtco/tokenqrusherJalankan perintah CLI-nya, atau tempel prompt ke Claude Code, Cursor, Codex, atau coding agent apa pun yang Anda pakai.