LEO Optima is a high-performance, self-hosted orchestration engine that slashes LLM API costs by 60-80% while simultaneously improving response quality. Designed for developers and enterprises who demand efficiency without compromise.

Built by Bader Jamal at Kadropic Labs.

In the current AI landscape, tokens are the new gold. Most applications bleed money through redundant queries, bloated prompts, and over-provisioned models. LEO Optima acts as a Smart Financial Layer for your AI stack.

Are you running OpenClaw (formerly Moltbot)? Autonomous agents like OpenClaw are incredible but notorious for "burning" tokens through recursive loops and repetitive state-checks.

LEO Optima is the perfect companion for your OpenClaw installation.

• Loop Deduplication: Prevents paying for the same state-check queries during agent loops.
• Context Slimming: Slashes the cost of long-running agent conversations by stripping redundant system prompts.
• Cost Guardrails: Monitor exactly how much your OpenClaw agent is spending in real-time.

Note: For the ultimate autonomous experience, Install LEO Optima alongside your OpenClaw setup and point your OpenClaw OPENAI_BASE_URL to LEO.

• Drastic Cost Reduction: Automatically saves up to 80% on OpenAI, Anthropic, and Gemini bills.
• Sub-Millisecond Speed: Serve repeated or semantically similar queries instantly from your local cache.
• Provider Agnostic: One unified interface for GPT-5, Claude 4.5, Gemini, and local models.
• Verifiable Truth: Built-in Byzantine Consensus ensures you get the most accurate answer every time.

LEO Optima isn't just a proxy; it's a multi-stage intelligence pipeline:

1. Request Deduplication: Identical concurrent requests are batched. Process once, serve all—zero extra cost.
2. Adaptive Semantic Cache: Powered by Johnson-Lindenstrauss Projection, LEO understands the intent of your queries, serving cached answers even if the wording changes.
3. Query Decomposition: Complex tasks are broken into atomic, cacheable fragments to maximize future reuse.
4. Prompt Slimming: Automatically removes "token fluff" from your prompts before they hit the paid API.
5. Byzantine Verification: Cross-verifies high-stakes queries across multiple models for absolute reliability.

Don't just take our word for it. Monitor your ROI in real-time with the built-in Kadropic Analytics Dashboard:

• Live USD Savings: Track every cent saved from hitting the paid APIs.
• Cache Performance: Visual hit-rate metrics and optimization ratios.
• Route Intelligence: See how LEO routes your traffic between Cache, Fast, and Consensus paths.

# Clone the core engine
git clone https://github.com/BADJAB22/leo-optima.git
cd leo-optima

# Configure your secrets
cp .env.example .env
# Add your Provider Keys (OpenAI, Anthropic, etc.) and your secret LEO_API_KEY

# Ignite
docker compose up --build -d

• Dashboard: http://localhost:3000
• API Proxy: http://localhost:8000

Simply update your OpenClaw .env or environment variables:

OPENAI_BASE_URL=http://localhost:8000/v1
# Ensure your LEO_API_KEY is passed in headers if required by your setup

LEO Optima is a seamless replacement for your existing OpenAI SDK setup. Change one line of code, and you're saving money.

from openai import OpenAI

# Simply point to your local LEO instance
client = OpenAI(
    api_key="your-provider-key", 
    base_url="http://localhost:8000/v1",
    default_headers={"X-API-Key": "your_secret_leo_key"}
)

# Use as normal—LEO handles the optimization magic
response = client.chat.completions.create(
    model="gpt-4o",
    messages=[{"role": "user", "content": "Analyze our Q4 growth strategy."}]
)

LEO Optima is 100% Free and Open Source. We believe high-performance AI should be accessible to every visionary developer.

• Found a bug? Open an Issue.
• Want to contribute? We love Pull Requests.
• Direct Contact? Connect with me on LinkedIn or X.

Crafted with precision by Kadropic Labs. License: MIT. Built for the builders.