Skip to content

MCP Server

Use the TestGen Model Context Protocol (MCP) server to give AI assistants access to your data quality results, profiling output, and test runs.

AI clients can investigate failures, summarize trends, look up source data, and trigger profiling or test runs through a structured set of tools.

What you can ask

Once connected, you can ask an AI assistant questions like:

  • "Which tests failed in our last test run?" — pulls failure summaries grouped by table, column, or test type.
  • "Show me sample rows that violate this test." — looks up the source data behind a failure.
  • "What's the root cause of the failures in the orders table?" — aggregates failures across runs and surfaces patterns.
  • "Which table is causing the most failures right now?" — ranks failures by table or column.
  • "Are failures getting better or worse?" — runs a time-series of test results.
  • "Compare the last two runs of this test suite — what regressed?" — diffs runs and highlights changes.
  • "What does the customer_id column look like?" — pulls column profile statistics, distributions, and quality scores.
  • "Profile the customer table group." — submits a profiling job and reports when it finishes.
  • "What does the Alpha Truncation test check?" — looks up test type definitions.

The exact phrasing does not matter — the assistant picks the right tool based on intent.

How it works

The MCP server exposes three kinds of capabilities:

  • Prompts — pre-built workflows you invoke directly through your AI client, typically as a slash command or quick-action menu item. They string together several tool calls for common investigations like Investigate failures or Compare runs. Exact UX varies by client.
  • Tools — operations the assistant calls during a conversation, picked based on what you ask (e.g., list_test_results, get_failure_summary, run_tests).
  • Resources — static reference documents the assistant can fetch by URI (e.g., the test type catalog, the glossary).

Tools and resources are picked automatically by the AI assistant based on what you ask — you do not have to specify them. Prompts are the exception — you invoke them yourself when you want a pre-built workflow. The complete catalog of prompts, tools, and resources is available in Supported Tools.

Get started

  1. Connect your AI client to the MCP endpoint at https://<your-testgen-host>/mcp.
  2. Authenticate with your TestGen account, either through the browser-based OAuth flow or an API token.
  3. Ask a question — start with a simple discovery prompt like "List the projects I have access to in TestGen."

See Set up the MCP Server for the full setup walkthrough.

Roadmap

The MCP server is fully supported, but the surface is still expanding. Expect additional tools, prompts, and workflow integrations in upcoming releases.