openclaw-master-skills/skills/open-optimise

🔋 Open-Optimise — Cost Optimizer for OpenClaw

Reduce your OpenClaw API costs by 70-97% through intelligent model routing, session management, output efficiency, and free model usage.

---

⚡ What Is This?

A complete cost optimization skill for OpenClaw — the open-source AI agent platform. It includes:

• 14-chapter agent playbook — Teaches your AI agent to route tasks to the cheapest adequate model
• 29 executable scripts — Audit, monitor, backup, health check, and report on your API spending
• 5 ready-to-use presets — One-command configs for different use cases and budgets
• Full user guide — With real example outputs from a live instance

The Problem

Every OpenClaw request sends ~140,000 tokens of system prompt overhead. On Claude Opus, that costs $0.71 per request. On free models, it costs $0.00. For simple questions, the answer quality is identical.

Most users run Opus/Sonnet for everything — including heartbeats (24/7 keep-alive pings), simple Q&A, and tool calls. This skill fixes that.

The Numbers

Setup	Monthly Cost (50 req/day)
All Opus (default)	~$1,065
All Sonnet	~$645
With this skill (budget preset)	~$30-60
With this skill (free preset)	~$0

---

📦 Installation

Quick Install

cd ~/.openclaw/workspace/skills/
git clone https://github.com/rajdeep09-dev/Open-Optimise.git cost-optimizer
chmod +x cost-optimizer/scripts/*.sh

Manual Install

1. Download the latest release
2. Extract to ~/.openclaw/workspace/skills/cost-optimizer/
3. Run chmod +x cost-optimizer/scripts/*.sh

Verify Installation

bash ~/.openclaw/workspace/skills/cost-optimizer/scripts/cost-audit.sh

You should see a full cost audit report.

First Run

Tell your OpenClaw agent:

/cost-setup

The skill walks through 6 setup steps, asking permission before each change:

1. Budget target
2. Free models
3. Default model
4. Heartbeat optimization
5. Session management
6. Response style

---

🔧 Requirements

Requirement	Details
OpenClaw	2026.3.x or later
Node.js	v18+ (ships with OpenClaw)
Shell	bash (Linux/macOS/WSL)
curl	For health checks and webhooks
Optional: OpenRouter API key	For free models ($0.00/request) — get one here
Optional: Webhook URL	For automated reports to Discord/Slack

---

📁 What's Included

cost-optimizer/
├── SKILL.md              # Agent playbook (14 chapters)
├── GUIDE.md              # User guide with real example outputs
├── README.md             # This file
├── VERSION               # 7.0.0
├── CHANGELOG.md          # Full version history
├── references/
│   ├── model-tiers.md    # Pricing reference card
│   └── setup-config.md   # Config templates
├── presets/
│   ├── solo-coder.preset.json
│   ├── writer.preset.json
│   ├── researcher.preset.json
│   ├── zero-budget.preset.json
│   └── agency-team.preset.json
└── scripts/              # 29 executable scripts
    ├── backup-config.sh
    ├── restore-config.sh
    ├── fallback-validator.sh
    ├── cost-audit.sh
    ├── heartbeat-cost.sh
    ├── cost-history.sh
    ├── session-replay.sh
    ├── provider-compare.sh
    ├── tool-audit.sh
    ├── token-counter.sh
    ├── context-monitor.sh
    ├── prompt-tracker.sh
    ├── compaction-log.sh
    ├── dedup-detector.sh
    ├── apply-preset.sh
    ├── token-enforcer.sh
    ├── config-diff.sh
    ├── idle-sleep.sh
    ├── setup-openrouter.sh
    ├── cost-monitor.sh
    ├── cost-dashboard.js
    ├── webhook-report.sh
    ├── cron-setup.sh
    ├── model-switcher.sh
    ├── provider-health.sh
    ├── model-test.sh
    ├── preset-manager.sh
    ├── multi-instance.sh
    └── parse-config.js

---

🛠️ Scripts Overview

🔴 Safety & Prevention

Script	What it does
backup-config.sh	Snapshot config before changes. Auto-prunes after 30 days.
restore-config.sh	List backups, pick one, restore, restart gateway. 5-second recovery.
fallback-validator.sh	Test every model in fallback chain with real API calls.

🔴 Where's The Money Going

Script	What it does
cost-audit.sh	Full config analysis → monthly estimate → actionable recommendations
heartbeat-cost.sh	Isolate heartbeat spend from real usage — the jaw-dropper
cost-history.sh	Recalculate real past usage across ALL model tiers
session-replay.sh	Exchange-by-exchange cost breakdown for any session
provider-compare.sh	Same model across providers — find cheapest route

🟡 Hidden Waste Detection

Script	What it does
tool-audit.sh	Find unused tools (overhead) + duplicate calls (execution waste)
token-counter.sh	Count system prompt overhead per file
context-monitor.sh	Track context growth, predict compaction
prompt-tracker.sh	Snapshot prompt size over time, detect silent growth
compaction-log.sh	Track compaction and memory flush events
dedup-detector.sh	Find redundant requests and tool calls

🟡 Active Optimization

Script	What it does
apply-preset.sh	One-command presets (auto-backs-up first)
token-enforcer.sh	Hard maxTokens limits: strict/moderate/normal/generous
config-diff.sh	Current vs recommended side-by-side with savings
idle-sleep.sh	Extend heartbeat during idle, restore on wake
setup-openrouter.sh	Generate OpenRouter provider config from API key

🟢 Monitoring & Reporting

Script	What it does
cost-monitor.sh	Real-time log parsing + --live tail mode
cost-dashboard.js	HTML dashboard with cost cards and savings bars
webhook-report.sh	Send daily reports to Discord/Slack/any webhook
cron-setup.sh	Pre-built cron templates for automated monitoring

🟢 Model Management

Script	What it does
model-switcher.sh	All models: status + cost + strength at a glance
provider-health.sh	Test all models: UP/DOWN/SLOW + latency
model-test.sh	Basic reachability test

🟢 Distribution & Sharing

Script	What it does
preset-manager.sh	Export/import named config presets
multi-instance.sh	Aggregate costs across multiple OpenClaw instances

---

🎯 Presets

Apply with one command:

bash scripts/apply-preset.sh <preset> [--dry-run]

Preset	Target Budget	Primary Model	Best For
free	$0-5/mo	Free models (OpenRouter)	Maximum savings
budget	$5-30/mo	DeepSeek V3.2	Daily driver
balanced	$30-100/mo	Claude Sonnet	Quality + savings
quality	$100+/mo	Claude Opus	Minimal optimization

Community Presets

Preset	Description
solo-coder	DeepSeek for coding, Opus for architecture reviews
writer	Sonnet for prose, DeepSeek for research/outlines
researcher	Haiku for web searches, Sonnet for synthesis
zero-budget	Free models only. $0/month target.
agency-team	Budget primary, higher concurrency for multi-agent work

---

📊 Example Output

cost-audit.sh

Primary Model:     anthropic/claude-opus-4-6
Heartbeat Model:   deepseek/deepseek-v3.2
Heartbeat Every:   55m
Model Aliases:     13 configured

── Cost Estimates (50 requests/day) ──

  Primary (Opus): $0.71/req → $1065/month
  Heartbeats (DeepSeek every 55m): $31.4/month (26/day)

  TOTAL ESTIMATED: $1096/month

  ── Comparison ──
  All Opus (unoptimized): $1623/month
  Current config:         $1096/month
  Savings:                $526/month (32%)

── Recommendations ──

  🔴 HIGH: Switch default from Opus to DeepSeek/MiniMax (~$1000/mo saved)
  🟡 MED: Add OpenRouter provider for free model access

provider-health.sh

  MODEL                         ALIAS       STATUS          LATENCY   COST
  ──────────────────────────────────────────────────────────────────────────
  Claude Opus 4.6               opus        ✅ AUTH OK       132ms     $0.71
  Claude Sonnet 4.6             sonnet      ✅ AUTH OK       132ms     $0.53
  Claude Haiku 4.5              haiku       ✅ AUTH OK       132ms     $0.15
  GPT-5.2                       gpt         ✅ AUTH OK       261ms     $0.44
  Gemini Flash                  flash       ✅ AUTH OK       84ms      $0.04
  DeepSeek V3.2                 deepseek    ✅ AUTH OK       100ms     $0.04

  Summary: 13 UP, 0 SLOW, 0 DOWN

fallback-validator.sh

  Testing 3 model(s) in fallback order...

  PRIMARY     anthropic/claude-opus-4-6      ✅ UP (2362ms)
  FALLBACK 1  anthropic/claude-sonnet-4-6    ✅ UP (3045ms)
  FALLBACK 2  anthropic/claude-haiku-4-5     ✅ UP (533ms)

  ✅ All 3 models in the chain are reachable.

See GUIDE.md for example outputs from every script.

---

🔄 Typical Workflows

"I just installed OpenClaw and want to save money"

bash scripts/cost-audit.sh                    # See current spend
bash scripts/apply-preset.sh budget --dry-run # Preview changes
bash scripts/fallback-validator.sh            # Verify chain works
# Then tell your agent: /cost-setup

"Something broke after a config change"

bash scripts/restore-config.sh               # List backups
bash scripts/restore-config.sh latest         # Restore last known-good

"Where is my money going?"

bash scripts/cost-audit.sh                    # Config overview
bash scripts/heartbeat-cost.sh                # Heartbeat spending
bash scripts/cost-history.sh                  # What-if across tiers
bash scripts/tool-audit.sh                    # Unused tool waste
bash scripts/token-counter.sh                 # System prompt bloat

---

📝 Agent Commands

When the skill is active, your agent responds to:

Command	Action
/cost-setup	Interactive setup wizard
/cost-status	Current model, session stats, estimated costs
/cost-audit	Full configuration audit
/free	Switch to free-only mode
/free off	Return to smart routing
/model <alias>	Switch to specific model

---

🤝 Contributing

1. Fork this repo
2. Create your feature branch (git checkout -b feature/my-optimization)
3. Test your scripts on a real OpenClaw instance
4. Commit your changes (git commit -am 'Add new optimization')
5. Push to the branch (git push origin feature/my-optimization)
6. Create a Pull Request

Adding a Preset

Create a .preset.json in presets/ following the existing format. Include a description of who the preset is for.

---

📄 License

MIT — Use it, modify it, share it. Save money.

---

🙏 Credits

Built for the OpenClaw community.

Stop paying $1,000/month for questions a free model can answer.