dremio

The Dremio MCP server provides a bridge between an LLM desktop client (like Claude) and a Dremio data environment using the Model Context Protocol (MCP). It enables the LLM to issue data-oriented requests to Dremio, manage authentication with PATs or OAuth, and route tool commands through a standardized MCP interface. The server supports a local development workflow (via uv) and production-ready deployment through Helm in Kubernetes, including streaming HTTP communication, Prometheus metrics, and TLS-enabled ingress. Tools exposed by the MCP server include configuration management (creating and listing server configs), runtime commands to start and test the MCP server, and dedicated coordination with Claude Desktop to keep the LLM in sync with Dremio configurations. You can start the server locally with uv run dremio-mcp-server and use the available subcommands to set up your Dremio and Claude configurations, verify connectivity, and then operate Claude to query Dremio data and manage datasets through the MCP interface.

Prerequisites:

• Python 3.11 or later
• uv package manager (https://docs.astral.sh/uv/getting-started/installation/)
• Access to a terminal/shell

1. Install uv and verify Python:

# Install uv (per the project instructions)
# Follow the uv install guide: https://docs.astral.sh/uv/guides/install-python/

# Ensure Python is 3.11+ installed
python --version

2. Clone or download the repository containing the Dremio MCP server.

3. Install dependencies (via uv):

# If your environment uses a virtual environment, activate it first
# Then you can install the MCP server dependencies as required by the repo (often via a setup or requirements file)
uv install  # if the project provides uv-based setup steps

4. Start the MCP server in a development mode (local workflow):

# From the top-level project directory
uv run dremio-mcp-server run

5. Verify the setup by listing and inspecting config options:

uv run dremio-mcp-server config list --type dremioai
uv run dremio-mcp-server config list --type claude

6. Create required configurations for Dremio and Claude as described in the repository’s Quick Start / Initial setup sections, for example:

# Create Dremio MCP server config
uv run dremio-mcp-server config create dremioai \
  --uri https://<dremio-uri> \
  --pat <dremio-pat> \
  # --project-id <dremio-project-id> (if applicable)

# Create Claude config (if using Claude Desktop)
uv run dremio-mcp-server config create claude

7. Start Claude Desktop and point it to the MCP server as configured in your Claude config file. You should now be able to issue natural language prompts to your LLM and have the MCP server route Dremio queries and data operations accordingly.

Tips and caveats:

• Use OAuth + External Token Provider for production deployments to securely manage access tokens.
• When testing locally, you can use a PAT for Dremio access; avoid printing the PAT to shell history by using file-based token references if supported by your setup.
• If deploying with Helm in Kubernetes, follow the Helm chart documentation and authentication guides for configuring Dremio URIs and credentials.
• The MCP server supports streaming HTTP mode for web deployments and exposes Prometheus metrics for observability.
• Ensure you configure the correct Dremio URI (cloud vs on-prem) and, if applicable, provide a project ID for cloud environments.
• For Claude Desktop integration, install Claude according to official docs and validate that the Claude config references the MCP server endpoints created by the dremio-mcp-server config setup.
• If you encounter issues with config loading, verify the default config paths shown in the Quick Start sections and ensure the config files exist at those locations.