Build AI agents with Exasol

Learn how to build AI agents in Exasol with agent skills and MCP server building blocks.

AI agents are more useful when they understand your data. Exasol provides two building blocks that bring AI to your data in a controlled, governed way: agent skills (knowledge) and the MCP server (access). Together, they support two workflows:

Build on Exasol

Use AI coding assistants like Claude Code or OpenAI Codex to write Exasol SQL, build UDFs, and manage your database with agents that understand Exasol’s dialect and can query your live schema.

Build with Exasol

Use Exasol as the data platform behind agentic solutions: agents that answer business questions, enrich metadata, or automate analytical workflows, all grounded in your database.

Skills and MCP: knowledge + access

Think of it like a textbook and a lab.

Agent skills are the textbook. They teach your AI assistant how Exasol works — SQL syntax quirks, UDF patterns, BucketFS conventions. Skills are static knowledge files that the assistant reads at the start of a conversation. With skills installed, your assistant writes correct Exasol-specific code instead of guessing based on generic SQL knowledge.

The MCP server is the lab. It gives your AI assistant a live connection to your database — browsing schemas, inspecting tables, running queries. With MCP access, your assistant can see your actual data and test its work against your real environment.

Either one works on its own, but they are strongest together. An assistant with skills writes better Exasol SQL. An assistant with MCP access can explore your database. An assistant with both can inspect your schema and then write optimized, Exasol-correct queries against it.

What’s available today

Three published agent skills cover the core areas of Exasol development:

Skill What it teaches
exasol-database Queries, data loading, SQL quirks
exasol-udfs UDF development in Python, Java, Lua, R, and SLC packaging
exasol-bucketfs File management in BucketFS

Skills work with Claude Code and OpenAI Codex. The MCP server works with any MCP-compatible client (Claude Desktop, Claude Code, Cursor, Windsurf, and others).

Get started

Fastest path: Install skills and connect the MCP server, then start asking your AI assistant to work with your database.

  1. Install agent skills:

    Copy 
    curl -fsSL https://raw.githubusercontent.com/exasol-labs/exasol-agent-skills/main/install.sh | sh

  2. Connect the MCP server to your AI assistant — see Connect AI assistants (MCP Server) for setup.

  3. Ask your assistant something Exasol-specific. For example: Write a Python UDF that classifies customer feedback using a Hugging Face model or Show me the largest tables in my SALES schema.

Next steps