queryLABSCO SUMMARY
Schema, profiling and charts on a Teradata instance, with enterprise authentication already wired.
The four analysis tools are the unusual part: missing values, negatives, distinct counts and standard deviation answer the first questions about a column without anyone writing SQL for them. Connections retry with exponential backoff and every call carries a QueryBand, so what an agent did shows up in workload management afterwards.
What it is
A Teradata server with several authentication mechanisms, column profiling tools and an interactive chart view built on ECharts.
What you get
- Databases, tables and views listed, and a table's column names and types shown (list_db, list_tables, show_tables_details)
- SQL executed against the database and returned as plain tabular results
- The same query rendered as an interactive chart in the client, across 19 chart types (visualize_query)
- Column profiling without writing the SQL: NULL counts, negative-value counts, distinct category counts, mean and standard deviation (list_missing_values, list_negative_values, list_distinct_values, standard_deviation)
- TD2, LDAP, Kerberos and JWT authentication through Teradata LOGMECH, plus OAuth with Keycloak
- A per-tool QueryBand, so activity is traceable in Teradata workload management
Requirements
uv, and a Teradata connection string in the form teradatasql://user:password@host/database — set in the environment or given as an argument. Cloned and installed with uv sync.
Setup effort
Build from source — git clone https://github.com/arturborycki/mcp-teradata.git
USE CASES
01
Finding your way around an unfamiliar Teradata database
List the databases, then the tables, then the columns and types of the one that matters.
02
Checking data quality before you trust a column
Read the NULL counts and negative-value counts rather than assuming the column is clean.
03
Understanding how a column is distributed
Pull distinct category counts and the mean and standard deviation in two calls.
04
Showing a result to someone who does not read tables
Render the query as an interactive chart in the client instead of exporting it somewhere else.
05
Answering an ad-hoc question from production data
Find the table, then run the query and chart the answer in the same conversation.
9 tools
Listed in project docs. Your client may show a different list.
visualize_querylist_dblist_tablesshow_tables_detailslist_missing_valueslist_negative_valueslist_distinct_valuesstandard_deviationBEFORE YOU INSTALL
What to know first
Tools you get
9 tools named in the project's documentation.
Requires
DATABASE_URI in your environment
Where it runs
The documented client runs the server on your machine.
INSTALLATION
Connect your client
Server setup
TERMINAL
git clone https://github.com/arturborycki/mcp-teradata.git
cd mcp-teradata
uv syncMCP Clients
Claude Desktop
CLAUDE DESKTOP · JSON
{
"mcpServers": {
"teradata": {
"command": "uv",
"args": [
"--directory",
"<YOUR_PATH_TO_MCP_TERADATA>",
"run",
"teradata-mcp"
],
"env": {
"DATABASE_URI": "teradatasql://<YOUR_DB_USER>:<YOUR_DB_PASSWORD>@<YOUR_DB_HOST>/<YOUR_DB_NAME>"
}
}
}
}
