Show HN: We launched hosted MCP and RAG (no infra, we host)

Hey HN,

We just put a fully managed Hosted MCP Server in front of our production-grade RAG stack. It solves the two things (will talk about that at end) that kept biting us (and most dev teams) when wiring agents to private data.

How does it work? (# 30-second flow)

→ CustomGPT Console → Deploy → MCP → Enable → Grab the generated endpoint + JSON schema → Add to MCP aware client

Point any MCP-aware tool (e.g. dozens of Agentic AI and workflow tools like n8n and Zapier; IDEs like Cursor; ChatGPT w/ MCP plugin, Anthropic’s Claude, etc.) at the endpoint.

So it's basically bringing RAG to MCP.

Back to the 2 things I talked about above are:

1. Agent Answer Accuracy a.k.a RAG accuracy – we benchmark at the top of public leaderboards for “business-doc” retrieval & no hallucination.

2. Ops drag – no k8s, patch cycles, or 3 a.m. TLS renewals. We host, autoscale, and watch the graphs.

Included in every CustomGPT.ai plan (free-trial friendly). Happy to share perf metrics or answer architecture questions.

Ask me anything!