"Bridging the gap between Unstructured Web and LLM Context."
Technical Co-founder and Senior Lead Architect specializing in high-performance serverless ecosystems. I build "Agent-Commerce-OS", a production-ready BaaS (Backend-as-a-Service) that allows AI agents to securely access, format, and ingest structured web data via standardized protocols.
🚀 Flagship Project: Project GHOST SHIP
The Secure MCP Gateway for Autonomous Data Retrieval.
Project GHOST SHIP serves as a middleware layer that normalizes external web content into RAG-optimized formats (Markdown/JSON) for LLMs.
- Layer A: The Edge Gateway (agent-commerce-gateway)
- Low-latency SSE/MCP aggregation on Cloudflare Workers.
- Security & Governance: Request sanitization, schema validation, and rate-limiting via Polar.sh.
- Layer B: The Normalization Engine (agent-commerce-core)
- Semantic Extraction: Python 3.12+ environment on Google Cloud Run for parsing complex HTML.
- LLM-Native Output: Converts raw documentation into clean Markdown using semantic parsing logic.
- Layer C: Audit & Cache (Supabase)
- Structured logging of tool execution results and ephemeral cache storage.
- 👉 Polar.sh (API Quotas & Entitlements) - Primary Governance Provider
- 👉 Project Documentation (llms.txt)
| Component | Protocol | Role & Description |
|---|---|---|
| GHOST-SHIP-GATEWAY | MCP / SSE | [Edge] Secure proxy for handling tool execution requests and schema validation. |
| AGENT-COMMERCE-CORE | REST / JSON | [Core] Heavy-lifting engine for web-to-markdown conversion and data structuring. |
| SEMANTIC-EXTRACTOR | Python SDK | [Lib] Domain-agnostic scraper logic for technical documentation. |
Optimized for AI Agents (AEO - Agent Engine Optimization).
- llms.txt : Context definitions for LLM crawlers.
- mcp.json : Dynamic tool definitions and endpoint discovery for MCP clients.
- Zero-Trust Logic: Strict input validation at the edge; execution in isolated environments.
- Semantic Naming: Every variable and endpoint is optimized for LLM self-documentation.
- Schema Compliance: Strict adherence to Pydantic v3 schemas to prevent hallucinations.
- Data Neutrality: We provide the pipe, not the advice. Pure technical infrastructure.
© 2026 Sakutto Works - Building the Semantic Web for Agents