Browse categories
Explore
Fiverr Pro
English
$
USD
I will build a secure snowflake or dbt mcp server for enterprise ai agents
Hariharan S
About this gig
Your data team is drowning in context-switching analysts jumping between Snowflake, dbt, Tableau, and Slack just to answer one business question. Meanwhile, your AI tools (Cursor, Claude, ChatGPT) can't securely touch your warehouse at all.
I build the missing layer: production-grade Model Context Protocol (MCP) servers that let AI agents query your Snowflake data and dbt models directly securely, governed, and without copy-pasting SQL into a chatbot.
What I deliver:
- Snowflake-managed MCP servers wired to Cortex Analyst (semantic NL-to-SQL), Cortex Search, and scoped SQL execution
- dbt MCP server setup for AI-assisted model development and metadata discovery
- Role-based access so agents never see data they shouldn't
- Direct integration with Cursor, Claude, and Slack your team asks questions in plain English, agents return governed answers
This isn't a demo. It's infrastructure your compliance team will approve and your engineers will actually trust.
If you want your data warehouse "AI-ready" without opening a security hole, message me with your current stack and I'll scope the fastest path to a working agent
Get to know Hariharan S
Hariharan S
Snowflake, dbt, and Agentic AI for Enterprise Data Stacks
- FromIndia
- Member sinceJan 2022
- Avg. response time1 hour
- Last delivery3 years
Languages
Tamil, English
Analytics Engineer specializing in modern data stack architecture — Snowflake, dbt Core, and Agentic AI via the Model Context Protocol (MCP).
I help data teams cut context-switching and Snowflake credit spend, and make enterprise data accessible through secure AI agents (Cursor, Claude, Slack) instead of manual SQL requests.
My work spans production dbt pipelines, Cortex Analyst semantic layers, dbt/Snowflake MCP server setups, and automated cost optimization tooling in Python/SQL.
If your stack needs to be fast, governed, and AI-ready — let's build it right.
FAQ
Is this secure enough for enterprise compliance requirements?
Yes. Every setup uses role-based access control at the Snowflake layer, so the MCP server and any connected AI agent only sees data explicitly authorized. Nothing bypasses your existing RBAC or masking policies. Self-hosted MCP via Snowpark Container Services is available if compliance requires it.
What's the difference between the Snowflake-managed MCP server and the dbt MCP server?
The Snowflake MCP server exposes Cortex Analyst, Cortex Search, and SQL execution so agents can query your warehouse directly. The dbt MCP server exposes your dbt project's metadata and lineage so AI coding assistants validate transformations against your actual models. Most clients want both.
Which AI clients can actually connect to this?
Any MCP-compatible client: Cursor, Claude Desktop, Claude Code, and Slack (via a Slack app wrapper). If your team uses a different tool, message me before ordering and I'll confirm compatibility before you commit.
Do I need a semantic layer already built before we start?
No. If you don't have semantic views yet, I'll scope and build a minimal one as part of the Standard/Premium packages. If you already have dbt models or Cortex semantic views, I'll build on top of them, which speeds up delivery significantly.
What happens after the agent is connected — is it just a chatbot?
No, it's closer to an internal API for your data, accessed via natural language. Once connected, stakeholders can ask questions in Slack or Cursor and get governed answers pulled live from Snowflake, without anyone hand-writing SQL.
