README.md
Save 30-75% on Manus AI credits with zero quality loss.
Credit Optimizer v5 for Manus AI
Save 30-75% on Manus AI credits with zero quality loss. ~55% average savings. Audited across 53 adversarial scenarios, 200+ tasks verified. Works as MCP server (free) or native Manus Skill ($9).
Available on: PyPI · MCP Registry · Smithery · GitHub · Landing Page
The Problem
Manus AI charges credits per task. Most users waste 30-75% of their credits because:
- Simple tasks run in Max mode when Standard would produce identical results
- Prompts contain redundant context that inflates token usage
- Tasks that could be batched are executed one by one
- Output formats are not optimized for the task type
Credit Optimizer fixes all of this automatically.
How It Works
Your Prompt
│
▼
┌──────────────────────────────────────────┐
│ Credit Optimizer v5 │
│ │
│ 1. Intent Classification (12 categories)│
│ 2. Complexity Scoring │
│ 3. Model Routing (Standard vs Max) │
│ 4. Prompt Compression │
│ 5. Batch Detection │
│ 6. Context Hygiene │
│ 7. Output Format Optimization │
│ │
│ Result: Optimized strategy + savings % │
└──────────────────────────────────────────┘
│
▼
Same quality output, fewer credits
Demo
> analyze_prompt("Build me a React dashboard with charts, auth, and database backend")
╔══════════════════════════════════════════════════════════════╗
║ CREDIT OPTIMIZER v5 — Analysis Report ║
╠══════════════════════════════════════════════════════════════╣
║ ║
║ Intent: code_generation (complex, multi-component) ║
║ Model: Max mode ✓ (correct for this complexity) ║
║ Savings: 35-45% estimated ║
║ Quality: 0% loss ║
║ ║
║ Strategy: Split into 3 sequential tasks ║
║ ┌──────────────────────────────────────────────────────┐ ║
║ │ Task 1: Database schema + API routes (Standard) │ ║
║ │ Task 2: Authentication flow (Standard) │ ║
║ │ Task 3: React dashboard + charts (Max) │ ║
║ └──────────────────────────────────────────────────────┘ ║
║ ║
║ Optimizations applied: ║
║ ✓ Model routing: Tasks 1-2 downgraded to Standard ║
║ ✓ Batch detection: 3 focused tasks vs 1 monolithic ║
║ ✓ Context hygiene: Removed redundant specifications ║
║ ✓ Output format: Structured code blocks per component ║
║ ║
╚══════════════════════════════════════════════════════════════╝
> analyze_prompt("Translate this paragraph to Spanish")
╔══════════════════════════════════════════════════════════════╗
║ CREDIT OPTIMIZER v5 — Analysis Report ║
╠══════════════════════════════════════════════════════════════╣
║ ║
║ Intent: translation (simple) ║
║ Model: Standard mode ✓ (Max unnecessary) ║
║ Savings: 60-70% estimated ║
║ Quality: 0% loss ║
║ ║
║ Recommendation: Use Standard mode ║
║ Translation tasks produce identical quality in Standard. ║
║ No splitting needed — single atomic task. ║
║ ║
╚══════════════════════════════════════════════════════════════╝
Real Results
| Metric | Val
Tools (1)
analyze_promptAnalyzes a prompt to provide an optimized strategy, model routing, and estimated credit savings.Configuration
claude_desktop_config.json
{"mcpServers": {"credit-optimizer": {"command": "uvx", "args": ["mcp-credit-optimizer"]}}}Try it
→Analyze my prompt: 'Build me a React dashboard with charts, auth, and database backend' to see how I can save credits.
→How can I optimize my request to 'Translate this paragraph to Spanish' to use fewer credits?
→Check if my complex coding task can be split into smaller, more efficient tasks to reduce Manus AI costs.