Preset MCP Server Overview

Overview

What it is:

The Preset MCP (Model Context Protocol) Server allows AI assistants like Claude and ChatGPT to interact directly with your Preset workspaces. Using natural language, users can explore dashboards, query datasets, execute SQL, create charts etc. without leaving their AI client of choice. MCP bridges your existing Preset data and permissions with the AI tools your team already uses, enabling faster, more intuitive access to business intelligence.

Who it’s for:

• Chart and dashboard creators can generate and iterate on visualizations faster by describing what they want.
• Viewers can ask questions about their data and get answers directly, without needing to request help from their data team.
• For all users, MCP can help surface patterns and insights you might not have thought to look for and explain results in plain language so findings are easier to understand and share.

Key value:

• Query data, explore dashboards, and generate charts through AI agents using natural language
• Quickly find the right datasets, charts, and dashboards for what you're trying to analyze
• Surface trends and insights in your data
• Respects your existing Preset roles and permissions, so no separate access management is needed

Authentication & Permission Management

For more details see: Preset MCP Server Authentication.

• Authentication methods:

  • OAuth (Recommended) — The simplest method for Claude and ChatGPT users. Authentication is handled automatically by redirecting you to Preset to log in. No tokens to manage or refresh.
  • JWT via API Token — Generate a short-lived access token by exchanging your Preset API credentials. Best for automation, CLI tools, and scripted setups.
  • Bearer Token (Direct) — Paste an existing JWT or browser session cookie directly. Useful for quick testing but not recommended for long-term use as tokens expire.

• Identity mapping:

  Users authenticate with their existing Preset credentials. Access is tied to the authenticated user's Preset account and workspace membership, so no separate identity setup is required.

• Permission model:

  All MCP operations respect the authenticated user's existing Preset roles, permissions, and data access.

Admin Controls & Restrictions

• Access controls:

  To get MCP enabled for your workspace, reach out to your customer success manager. Once enabled, any user in the workspace can connect to it and get started.

• Scope restrictions:

  MCP access follows your existing workspace configuration. Users can only interact with the datasets, charts, and dashboards they already have permission to access in Preset.

• Rate limits & usage controls:

  There are no additional rate limits or usage limits applied on the Preset side. Usage limits will be subject to limits of your AI client.

• Audit & visibility:

  Actions taken via MCP are captured in Preset's standard audit logs. Currently, logs do not differentiate between actions taken via MCP and those taken directly in the product.

Available MCP Tools

System & Health

• health_check — Test connectivity and confirm service is running
• get_instance_info — Get instance statistics, counts, and activity metrics

Discovery & Schema

• get_schema — Get comprehensive schema metadata for charts, datasets, or dashboards
• list_datasets — List datasets with filtering and search
• list_charts — List charts with filtering and search
• list_dashboards — List dashboards with filtering and search

Read Details

• get_dataset_info — Get dataset metadata (columns, metrics, schema)
• get_chart_info — Get chart metadata by ID or UUID
• get_chart_data — Get the actual data behind a chart
• get_chart_preview — Get chart preview in various formats (URL, ASCII, table, vega_lite)
• get_dashboard_info — Get dashboard metadata by ID, UUID, or slug

Chart Creation & Editing

• generate_chart — Create a chart preview, optionally save permanently
  • Current supported chart types include line, bar, area, scatterplots, tables, and interactive table charts.
• generate_explore_link — Generate an interactive explore URL (preferred for visualization)
• update_chart — Update an existing saved chart
• update_chart_preview — Update a cached preview without saving

Dashboard Management

• generate_dashboard — Create a new dashboard from chart IDs
• add_chart_to_existing_dashboard — Add a chart to an existing dashboard

SQL & Data Execution

• execute_sql — Execute SQL queries against a database
• open_sql_lab_with_context — Open SQL Lab with pre-populated query and context

Proven AI MCP Workflows

Data discovery
Describe what you're looking for and the AI will find the relevant datasets, charts, and dashboards in your workspace and link you directly to them.

Ad-hoc analysis and chart creation
Ask business questions in plain language and the AI will query your data, generate charts, and return results directly in your Claude or ChatGPT conversation. You can filter by specific values, text patterns, or lists, ask follow-up questions, and iterate on the analysis without leaving your AI client.

Current supported chart types include line, bar, area, scatterplots, tables, and interactive table charts.

Insight generation
Ask the AI what's interesting or unusual in your data and it will surface trends, patterns, or outliers you might not have thought to look for, with explanations in plain language.

SQL generation and execution
Describe what you want to analyze and the AI will write and run the SQL against your databases, or generate a link to open SQL Lab with the query pre-populated so you can review and refine it yourself.

Current Limitations

The following actions in Preset are not currently supported via MCP:

• Creating, editing, or deleting database connections
• Creating, editing, or deleting datasets
• Uploading data
• Deleting charts or dashboards
• Editing dashboard layouts or filters
• Creating, editing, or managing alerts and reports
• Managing users, roles, or permissions
• Managing saved queries in SQL Lab

Chart creation is limited to the following chart types:

• Line, bar, area, and scatter charts
• Interactive tables and tables

Other chart types such as pie charts, maps, pivot tables, and big number charts are not supported