Configure a Local Tableau MCP Service
This page is a guide for individual users running Tableau MCP locally — as a stdio server
launched by your AI tool (MCP client) via npx, node, or Docker. It explains where to put
configuration and which settings are most useful when you are running the server just for yourself.
If you have not set up a local server yet, start with Getting Started. For an administrator-focused view of a shared, multi-user deployment, see Configuration for Site Administrators.
A "local service" here means the default stdio transport, where your MCP
client starts and stops the server process for you. Configuration is supplied through environment
variables, most commonly in your client's MCP config env block or in a .env / env.list file.
Where to put configuration
For a local server, you can provide environment variables in any of these ways:
- In your MCP client config — the
envblock of the server entry (see the examples below). This keeps everything self-contained in your client's configuration. - In a
.envfile — placed where the server can read it. Useful when you build from source and want to keep credentials out of your client config. - In an
env.listfile — passed to Docker with--env-file env.list. - System / user environment variables — inherited by the server process.
Treat your .env, env.list, and client config as secrets. They can contain a PAT value, OAuth
secrets, or connected-app keys. Do not commit them to source control. Credentials are masked in
logs by default — leave DISABLE_LOG_MASKING at false.
Minimal configuration
At a minimum, a local server needs to know how to reach Tableau and how to authenticate. For local, single-user use, a Personal Access Token (PAT) is the simplest option.
{
"mcpServers": {
"tableau": {
"command": "npx",
"args": ["-y", "@tableau/mcp-server@latest"],
"env": {
"SERVER": "https://my-tableau-server.com",
"SITE_NAME": "my_site",
"AUTH": "pat",
"PAT_NAME": "my_pat",
"PAT_VALUE": "pat_value"
}
}
}
}
| Variable | Purpose |
|---|---|
SERVER | URL of your Tableau Cloud pod or Tableau Server. |
SITE_NAME | Content URL of your site. |
AUTH | Authentication method — pat is typical for local use. |
PAT_NAME / PAT_VALUE | Your personal access token credentials. |
SITE_NAME is the site's Content URL, which can differ from the display name. For Tableau
Server, you can leave it blank to use the default site.
Configuring behavior
Once the basics work, you can tailor the server to your workflow. These variables all work the same
way locally as they do server-side — set them in your env block or .env file.
Choose which tools are available
INCLUDE_TOOLS— register only the listed tools or tool groups.EXCLUDE_TOOLS— register everything except the listed tools.- Provide only one of the two.
"env": {
"INCLUDE_TOOLS": "query-datasource,list-datasources,get-datasource-metadata"
}
Some tool groups are gated behind their own flags and must be turned on explicitly:
FLOW_READ_TOOLS_ENABLED— enables the Tableau Prep flow tools.ADMIN_TOOLS_ENABLED— enables admin-only tools. These still require your Tableau user to have a site administrator role; enabling the flag alone does not grant access.
Limit which content the tools touch
If you only ever work with a handful of data sources, workbooks, or projects, you can scope the server to them. See Tool Scoping.
"env": {
"INCLUDE_DATASOURCE_IDS": "2d935df8-fe7e-4fd8-bb14-35eb4ba31d4"
}
Cap result sizes
Keep responses small and fast (and easier for your model to reason over) with result limits:
MAX_RESULT_LIMIT— a global cap for tools with alimitparameter.MAX_RESULT_LIMITS— per-tool caps.
"env": {
"MAX_RESULT_LIMIT": "50",
"MAX_RESULT_LIMITS": "query-datasource:1000,list-datasources:20"
}
Tune logging for troubleshooting
When something is not working, local logs are your best friend:
ENABLED_LOGGERS— addfileLoggerto write detailed logs (including tool calls and REST API traces) to disk.FILE_LOGGER_DIRECTORY— where those log files go.LOG_LEVEL— set todebugfor maximum detail.DEFAULT_NOTIFICATION_LEVEL— the minimum level of notifications sent to your MCP client.
"env": {
"ENABLED_LOGGERS": "appLogger,fileLogger",
"LOG_LEVEL": "debug"
}
For local debugging you can also use the MCP Inspector to call tools directly and watch requests and responses. To step through the server with breakpoints, see Debugging.
Data source credentials
If you query data sources that require their own username and password, provide them with
DATASOURCE_CREDENTIALS.
Running from source or Docker
The configuration above applies regardless of how you launch the server — only where you put the environment variables changes:
- From source (
node build/index.js) — keep your settings in a.envfile next to the build instead of the client config'senvblock. - Docker — pass them with
--env-file env.list.
For the full launch snippets (npx, Node, and Docker) and first-time setup, see Getting Started; for details on the container image, see Docker.
Applying changes
Environment variables are read when the server process starts. After editing your client config,
.env, or env.list, relaunch your AI tool (or refresh its MCP tools) so a fresh server
process picks up the new values. When building from source, rebuild with npm run build after
pulling changes before relaunching.
Related
- Getting Started — install and first run.
- Environment Variables — the full catalog of settings.
- Authentication — PAT, direct trust, and OAuth options.
- Tool Scoping — limit tools to specific content.
- MCP Inspector — test tools locally.
- Configuration for Site Administrators — the shared-deployment counterpart to this page.