Hey everyone,
I wanted to share an MCP server I've been building called SAPiola. It focuses on bridging the gap between heavy enterprise data and LLM agents without relying on slow, expensive batch ETL pipelines.
What it does under the hood:
- Zero-ETL Streaming: It consumes raw SAP HANA Change Data Capture (CDC) events in real-time, mapping complex relational schemas straight into a localized Knowledge Graph.
- High-Speed Embedded Engine: Built in Rust using Fjall (LSM trees) for highly concurrent relational state storage and localized Cypher query parsing. It keeps memory footprints incredibly low with sub-millisecond graph query responses.
- Proxy Pointer Hybrid RAG: The orchestration layer uses a proxy-pointer approach. It resolves domain constraints (with strict RBAC checks) and graph indices before executing RAG, helping eliminate LLM hallucinations on deeply nested corporate data.
- Native MCP Capabilities: Exposes tools for LLM agents (like Claude Desktop or Cursor) to natively trigger adversarial BFS traversals, fetch vertex properties, or query SAP ERP records using natural language.
I would love to have this added to the community servers list so developers building enterprise-grade agentic tools have an efficient way to pull context out of SAP environments.
Let me know if you need any adjustments or if you have any questions about the Rust-embedded architecture!
Hey everyone,
I wanted to share an MCP server I've been building called SAPiola. It focuses on bridging the gap between heavy enterprise data and LLM agents without relying on slow, expensive batch ETL pipelines.
What it does under the hood:
I would love to have this added to the community servers list so developers building enterprise-grade agentic tools have an efficient way to pull context out of SAP environments.
Let me know if you need any adjustments or if you have any questions about the Rust-embedded architecture!