mcpField
T
token-optimizer

token-optimizer

github.com/smartpeopleconnectedAgent workflows771 installsUpdated 36w ago
Repository

About token-optimizer

token-optimizer is a command-line skill for reducing OpenClaw token usage through model routing, heartbeat provider choices, session context management, and caching. It also offers configuration previews, backups, rollback, health checks, and savings reports.

Key features

  1. 01Route tasks across model tiers
  2. 02Use multiple heartbeat providers or disable heartbeats
  3. 03Reduce session context size
  4. 04Reuse prompts through caching
  5. 05Preview changes and restore config backups
  6. 06Run health checks and savings reports

Use cases

  1. 01Tune OpenClaw model selection for tasks
  2. 02Route heartbeats to Ollama, LM Studio, or Groq
  3. 03Preview configuration edits before applying
  4. 04Restore a previous OpenClaw configuration

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/smartpeopleconnected/openclaw-token-optimizer --skill token-optimizer

After installing, type/token-optimizerin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

.clawhubignore
_meta.json
setup.py
.gitignore
README.md
LICENSE
CHANGELOG.md
cli.py
requirements.txt
skill.json
.github/
FUNDING.yml
src/
__init__.py
colors.py
analyzer.py
optimizer.py
verify.py
templates/
openclaw-config-optimized.json
OPTIMIZATION-RULES.md
USER.md
SOUL.md
test/
simulation_test.py

Details

From the project README

Token Optimizer for OpenClaw

Slash your AI costs from $1,500+/month to under $50/month.

The Problem

OpenClaw defaults prioritize capability over cost. You're burning expensive Sonnet/Opus tokens on tasks Haiku handles perfectly, paying for API heartbeats that could run free locally, and loading 50KB of context when 8KB suffices.

The Solution

Four core optimizations plus powerful tooling:

Model Routing (92% savings)

Haiku by default, Sonnet/Opus only when needed

Multi-Provider Heartbeats (100% savings)

Route heartbeats to Ollama, LM Studio, Groq, or disable entirely. Not locked to one provider.

Session Management (80% savings)

Load 8KB instead of 50KB context

Caching (90% savings)

Reuse prompts at 10% cost

New in v1.0.8
  • Rollback - List and restore config backups instantly
  • Health Check - Quick system status in one command
  • Diff Preview - See exactly what changes before applying
  • --no-color - CI/pipeline friendly output

Cost Comparison

PeriodBeforeAfter
Daily$2-3$0.10
Monthly$70-90$3-5
Yearly$800+$40-60

What's Included

  • One-command optimizer with diff preview
  • Multi-provider heartbeat (Ollama, LM Studio, Groq)
  • Config rollback and health check commands
  • Ready-to-use config templates
  • SOUL.md & USER.md templates
  • Optimization rules for agent prompts
  • Verification and savings reports

What This Tool Modifies

All changes are written under ~/.openclaw/. A backup is created before any modification.

PathPurpose
~/.openclaw/openclaw.jsonMain OpenClaw config (model routing, heartbeat, budgets)
~/.openclaw/backups/Timestamped config backups (created automatically)
~/.openclaw/workspace/Template files (SOUL.md, USER.md, IDENTITY.md)
~/.openclaw/prompts/Agent prompt optimization rules
~/.openclaw/token-optimizer-stats.jsonUsage stats for savings reports

Safe by default - All commands run in dry-run (preview) mode. Pass --apply to write changes.

Quick Start

# Install
clawhub install token-optimizer

# Analyze current setup
python cli.py analyze

# Preview changes (dry-run by default)
python cli.py optimize

# Apply all optimizations
python cli.py optimize --apply

# Verify setup
python cli.py verify

# Quick health check
python cli.py health

# Configure heartbeat provider (preview)
python cli.py setup-heartbeat --provider ollama

# Configure heartbeat provider (apply)
python cli.py setup-heartbeat --provider ollama --apply

# List and restore backups
python cli.py rollback --list
python cli.py rollback --to <backup-file>

Configuration Generated

{
  "agents": {
    "defaults": {
      "model": { "primary": "anthropic/claude-haiku-4-5" },
      "cache": { "enabled": true, "ttl": "5m" }
    }
  },
  "heartbeat": {
    "provider": "ollama",
    "model": "ollama/llama3.2:3b"
  },
  "budgets": {
    "daily": 5.00,
    "monthly": 200.00
  }
}

Links

Author

Smart People Connected

License

MIT License - Free to use, modify, and distribute.

5 minutes to setup. 97% cost reduction. Stop burning tokens. Start building.

Comments

Markdown supported