token-saver
This skill optimizes workspace .md files and provides a model-aware compaction/dashboard to reduce API costs. It reads home config ~/.openclaw/openclaw.json, scans sessions in ~/.openclaw/agents/main/sessions, and writes to AGENTS.md (adds ## 📝 Token Saver — Persistent Mode).
⚡ Token Saver v3
💡 Did you know? Every API call sends your workspace files (SOUL.md, USER.md, MEMORY.md, AGENTS.md, etc.) along with your message. These files cost real money on every message.
Token Saver v3 is model-aware — it knows your model's context window and adapts recommendations accordingly.
Quick Start
/optimize
That's it. You'll see a dashboard with your savings options.
What's New in v3
| Feature | v2 | v3 |
|---|---|---|
| Compaction presets | Fixed (80K/120K/160K) | Dynamic (% of model's context) |
| Model detection | Fragile, env-only | Robust fallback chain |
| Context windows | Not tracked | Full registry (9 models) |
| Savings estimates | Static | Model-aware pricing |
Dashboard Preview
╭─────────────────────────────────────────────────────────╮
│ ⚡ TOKEN SAVER v3 │
│ Reduce AI costs by optimizing what gets sent each call │
╰─────────────────────────────────────────────────────────╯
🤖 **Model:** Claude Opus 4.5 (200K context)
Detected: openclaw.json
📊 **Context Usage:** [████████░░░░░░░░░░░░] 42% (84K/200K)
📁 **WORKSPACE FILES** (sent every API call)
┌──────────────────────┬───────┬────────────────┐
│ File │ Tokens│ Can Save │
├──────────────────────┼───────┼────────────────┤
│ 🔴 AGENTS.md │ 1180 │ -825 (70%) │
│ 🟢 SOUL.md │ 235 │ ✓ optimized │
...
Commands
| Command | What It Does |
|---|---|
/optimize | Dashboard with files, models, context usage |
/optimize tokens | Compress workspace files (auto-backup) |
/optimize compaction | Chat compaction control (model-aware) |
/optimize compaction balanced | Apply balanced preset (60% of context) |
/optimize models | Detailed model audit with registry |
/optimize revert | Restore backups, disable persistent mode |
Dynamic Presets
Presets adapt to your model's context window:
| Preset | % | Claude 200K | GPT-4o 128K | Gemini 1M |
|---|---|---|---|---|
| Aggressive | 40% | 80K | 51K | 400K |
| Balanced | 60% | 120K | 77K | 600K |
| Conservative | 80% | 160K | 102K | 800K |
Model Registry
9 models with context windows and pricing:
- Claude: Opus 4.5, Sonnet 4, Haiku 3.5 (200K)
- Gemini: 2.0 Flash, 2.5 Pro (1M)
- OpenAI: GPT-4o, GPT-4o Mini (128K)
- Others: DeepSeek V3 (64K), Kimi K2.5 (128K)
File Compression
Before (verbose):
When Ruben greets me in the morning, I should proactively review our task list...
After (compressed):
MORNING: greeting → review(todos+pending+urgent)
Same meaning. 90% fewer tokens. Real savings.
Safety
- ✅ Auto-backup before any changes
- ✅ Smart bypass — skips already-optimized files
- ✅ One-command revert —
/optimize revert - ✅ No external calls — all local
Install
# From ClawHub
clawhub install token-saver --registry "https://www.clawhub.ai"
# Or clone directly
git clone https://github.com/RubenAQuispe/token-saver.git
License
MIT — Use it, modify it, share it.