Backend / AI Integration ยท 2025
Rust MCP server that gives LLM clients structured access to ERP analytics without exposing raw database access.
Developed an integration layer translating natural language queries from LLMs into secure, structured ERP analytics requests using MCP.
Exposing sensitive ERP data to AI without risking data leakage or hallucinated SQL injections, while adhering strictly to MCP standards.
Enforced strict parameter binding and predefined endpoints instead of raw SQL. Implemented a zero-trust model where AI only receives pre-authorized data blocks.
Self-hosted AI agent platform pairing a Rust/Axum backend with a Vue 3 workspace: multi-agent orchestration, hybrid RAG, MCP tools, workspace access, approval controls, and OpenAI-compatible APIs in one deployable stack.
OpenAI-compatible AI router in Rust that puts every upstream model behind one endpoint, with prefix routing, fallback combos, token saving, usage metering, and an embedded dashboard in a single binary.
Self-hosted deployment and operations control plane for Docker, Compose, Git builds, domains, and remote infrastructure.
Share your goals and constraints and I'll help map the scope and the technical approach.