Dremio MCP server
Dremio MCP server: an agent browses tables, looks for patterns in the data and analyzes cluster workload.
- 53 ★ on the repository
- Last commit
- Apache-2.0
Extended access
Why this level
- A personal Dremio token opens every source and table its owner can reach to the agent.
- The `allow_dml` option lets the server create views in Dremio.
Install
- Works in
Manual install
uv run dremio-mcp-server --helpSanity check after cloning `dremio/dremio-mcp`. `uv` and Python 3.11+ are needed.
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- A personal Dremio token opens every source and table its owner can reach to the agent.
- The `allow_dml` option lets the server create views in Dremio.
This is third-party code. Review the repository files before installing.
What it does
The server connects an agent to the Dremio data lake analytics platform. It runs in three modes that can be combined with commas. `FOR_DATA_PATTERNS` opens tables and data for pattern discovery. `FOR_SELF` lets the agent analyze the Dremio system itself, including workload analysis. `FOR_PROMETHEUS` connects Prometheus metrics when they are enabled in the cluster. Sign-in uses a personal access token (PAT). Dremio Cloud adds a project ID. Extra options switch on semantic search and allow creating views (`allow_dml`). For production Kubernetes deployments there is a Helm chart with streaming HTTP, OAuth and Prometheus metrics.
Who it is for. Data analysts and Dremio administrators who want to ask about tables and cluster load in plain words from Claude Desktop.
Good fit when
- There are many tables in the data lake and the agent has to point to where the needed fields live.
- Slow queries are analyzed against cluster load in `FOR_SELF` mode.
- Prometheus metrics of the cluster are read alongside Dremio data.
- A team deploys the server in Kubernetes through Helm with OAuth.
Not a fit when
- The data lives outside Dremio: the server knows only this platform.
- The agent must not hold a token: a PAT opens every source its owner can reach.
- Official support is required: the server is not covered by it.
Example request
Show which tables in the sales source relate to orders and find patterns in order totals by month.MCP
- Transport
- stdio, http
- Authentication
- API key
Limitations
Local installation runs from a cloned repository and the README names no registry package. Python 3.11 or newer and the `uv` manager are required. The server config is created with `uv run dremio-mcp-server config create dremioai` and Claude Desktop is configured by a separate command. The local setup is described for Claude Desktop only. The server is not part of Dremio product support. For Kubernetes you build the image yourself with `docker build -t dremio-mcp:0.1.0 .`.
How to disable. Delete the `Dremio` section from `claude_desktop_config.json` and the `~/.config/dremioai/config.yaml` file that holds the token. Revoke the token in Dremio settings. A Kubernetes deployment is removed with `helm uninstall`.
FAQ
How do I install the Dremio MCP server?
Clone the repository, install `uv` and check `uv run dremio-mcp-server --help`. Then create the config with `uv run dremio-mcp-server config create dremioai` using the URI and token, followed by `uv run dremio-mcp-server config create claude`.
How do I keep the token out of shell history?
Put the PAT in a file and pass the path with an `@` prefix: `--pat @/path/to/tokenfile`. The same prefix works in the `pat` field of `config.yaml`.
Which address goes for Dremio Cloud?
For the US region use `https://api.dremio.cloud` or the shorthand `prod`, for Europe `https://api.eu.dremio.cloud` or `prodemea`. A cloud project also needs `--project-id`.
What do the server modes do?
`FOR_DATA_PATTERNS` reads tables and data, `FOR_SELF` analyzes the Dremio system and load, `FOR_PROMETHEUS` pulls Prometheus metrics. List modes with commas.
How do I deploy the server for a team?
Use the Helm chart in `helm/dremio-mcp`. The recommended production option uses OAuth, and the README marks the PAT option as for testing.
Discussion
No questions or reviews yet.
Related
MCP serversOfficial Tableau MCP server: the agent finds data sources, sends queries to them and opens dashboards and Pulse metrics
One MCP server for SQLite, SQL Server, PostgreSQL and MySQL: read and change data, edit schemas and export to CSV and JSON
A DataHub MCP server: an agent searches tables and columns by metadata, traces lineage and finds analysts' SQL queries
A local Databricks API emulator with a 67-tool MCP server: agents run SQL and jobs without a real workspace
MySQL MCP Server Pro
mysql_mcp_server_pro
A MySQL MCP server that runs SQL under access roles, finds tables by comments and analyzes locks, indexes and query plans
Apache Doris's official read-only MCP server: catalogs, queries, cluster diagnostics, search and lineage
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| Dremio MCP server (this card) | yes | API key | 53 | |
| MCP Toolbox for Databases | yes | not required | 16,812 | connects an agent to a set of databases; Dremio MCP server works with the single Dremio platform and can analyze cluster load |
| DataHub MCP Server | yes | API key | 85 | searches the DataHub metadata catalog; Dremio MCP server reads tables and data inside Dremio itself |
README in short
The README splits installation into a remote Helm deployment and a local one for a desktop client. It then covers the settings file format, the three server modes, log locations per system and links to architecture, tools and settings docs.