doris_catalogLABSCO SUMMARY
Discovery first, then one exact call.
Calling a domain with an empty argument returns the children your route and privileges authorize together with their schemas, so an agent plans against what is callable rather than against a catalog it might not be allowed to touch. That also keeps eight domain tools in context instead of the full expansion. The built-in surface is read-only throughout — the admin domain is reserved and never registered — so the Doris account you connect as is the whole of the access decision.
What it is
The Apache Doris project's own MCP server. Tools are grouped into domains you call empty to discover which children your route and privileges actually authorize, then call again naming one child and its arguments — so the surface stays small and Doris RBAC remains the final authority over what comes back.
What you get
- doris_catalog — metadata navigation: catalogs, databases, tables, columns and their definitions
- doris_query — read-only querying and query diagnostics, including explain and profile work
- doris_cluster — runtime and resource state: nodes, workload groups, and what the cluster is currently doing
- doris_pipeline — ingestion and freshness, for answering whether the data you are about to query has actually landed
- doris_search — text, vector and hybrid retrieval against Doris's own indexes
- doris_governance — quality, lineage, audit and access, for questions about where a table came from and who touched it
- doris_lakehouse — external catalogs and semi-structured data, so a query can reach outside Doris's own storage
- doris_semantic — optional semantic consumers, registered only when you enable them
- An admin domain that is reserved and never registered, which is what makes the built-in surface read-only throughout
Requirements
Python 3.12 or later, and network access to a supported Doris FE — DORIS_HOST and DORIS_PORT for the MySQL endpoint, DORIS_USER and DORIS_PASSWORD, DORIS_DATABASE for the default, and DORIS_FE_HTTP_HOSTS for the tools that use the HTTP interface. Install with pip and run doris-mcp-server over stdio or Streamable HTTP; non-loopback HTTP requires authentication. One stdio gotcha the docs call out: nothing else may write to stdout, because that is where the MCP frames go. The Doris user you connect as decides what any tool can see.
Setup effort
One command plus credentials — pip install doris-mcp-server==1.0.0, then supply credentials
USE CASES
01
Answering a question about data you have not seen before
Navigate the catalog to the table, read its columns, then run the query.
02
A query is slower than it should be
Read its diagnostics alongside what the cluster and its workload groups were doing at the time.
03
Checking whether today's data has actually arrived
Look at ingestion and freshness before trusting the number you are about to report.
04
Working out where a table came from
Read its lineage and the audit trail on it.
05
Searching rather than filtering
Use text, vector or hybrid retrieval against the Doris indexes.
06
Joining Doris data with something outside it
Reach the external catalog through the lakehouse domain and query across both.
8 tools
Listed in project docs. Your client may show a different list.
doris_querydoris_clusterdoris_pipelinedoris_searchdoris_governancedoris_lakehousedoris_semanticBEFORE YOU INSTALL
What to know first
Tools you get
8 tools named in the project's documentation.
Requires
- doris-mcp-server on PATH
- Python 3.12 or later
- DORIS_PASSWORD in your environment
- DORIS_USER in your environment
INSTALLATION
Connect your client
Add the block to your client, or run the command directly.
Server setup
Configuration file
JSON
{
"mcpServers": {
"doris-stdio": {
"command": "doris-mcp-server",
"args": [
"--transport",
"stdio"
],
"env": {
"DORIS_HOST": "127.0.0.1",
"DORIS_PORT": "9030",
"DORIS_USER": "root",
"DORIS_PASSWORD": "<YOUR_DORIS_PASSWORD>"
}
}
}
}pip
TERMINAL
pip install doris-mcp-server==1.0.0
