Connecting Pigment's MCP Server to AI systems gives access to tools in two modes.
⚠️ Important
The default mode and advanced mode outlined below govern which tools are available to external AI systems. However, access to MCP tools does not supersede user access rights and permissions. Switching on MCP does not give users modeling rights that they don't have in Pigment.
For example:
If advanced mode is enabled, the AI system can see and use the tool "create_metric".
But if the user doesn't have the "Configure Blocks" permission, it just returns an error. Even if the AI can use the tool, it can't perform the action because the user doesn't have the permission.
Default mode
Default mode provides read-only tools for querying and exploring data. To make a Block's data accessible, you need to enable AI data access on it.
get_organization_info
get_organization_infoDescription | Sample Prompt |
|---|---|
Get information about the current organization (Workspace): ID and name. Call this to identify which organization (Workspace) the user belongs to. | “What Workspace am I in?” |
get_applications
get_applicationsDescription | Sample Prompt |
|---|---|
List all Pigment applications the user can access, with their ID, name, and description. Call this first to obtain application IDs required by other tools. | “List all my Pigment Applications.” |
get_ai_metrics
get_ai_metricsDescription | Sample Prompt |
|---|---|
List AI-enabled Metrics in an Application. A Metric holds values organized along Dimensions (e.g., month, region, product). Not all Metrics may be AI-enabled. Returns Metric IDs and names. Use to discover available data and obtain Metric IDs for get_metric_description and query_data. | “Show me Metrics available for analysis in the Financial Planning application.” |
get_metric_description
get_metric_descriptionDescription | Sample Prompt |
|---|---|
Get the structure of a Metric: name, data type, description, Dimensions (with sample Items), and Scenarios. Use to understand how a Metric is organized. | "Describe the Revenue Metric." "Show me the structure of the Headcount Metric." "What are the Dimension Items for Product in the Revenue Metric?" |
query_data
query_dataDescription | Sample Prompt |
|---|---|
Retrieve data from a single Metric using a natural-language query. Returns values and metadata (name, link, Dimensions, filters). One Metric per call. | “Show me Revenue.” “Show me Sales for top 10 products by Year.” |
Advanced mode
Advanced mode provides read and write tools for modeling and editing your Applications. For instance, advanced tools let you:
Create Applications
Build and edit Blocks (Dimensions, Metrics and Tables)
Write formulas
Design Views and Boards
Input data
Audit performance
⚠️ Important
We recommend using advanced mode with Skills (see section below).
Each user can enable advanced mode for their account in Settings > Advanced features > Enable advanced tools on MCP server. Advanced mode is off by default.
ℹ️ Note
Advanced mode includes a
searchtool that lets the AI systems scan and understand your entire Application logic. This helps the systems provide better answers. Thesearchtool is not subject to AI data access. It exposes all Blocks’ metadata (e.g. names, data types, Dimensions, formulas) to users, but no actual data is accessible through it. We recommend that you do not put any sensitive information in Block metadata as it is not subject to access rights. If you want to prevent a user from accessing Block metadata, you must remove that user’s access to the Block’s Application in Pigment.
Skills
Skills are domain-knowledge files that give AI systems the context they need to work effectively with Pigment. They package Pigment-specific knowledge — including formula syntax, modeling best practices, and application patterns — so the AI can produce accurate, high-quality results without requiring you to explain the context yourself.
We recommend using Skills when working in advanced mode. Without them, AI systems lack the Pigment-specific knowledge needed to model reliably.
Pigment Skills are available in this public repository and can be set up in two ways:
Plugin: For Claude Code and Cursor, use the dedicated plugins available in the repository. These bundle the MCP Server configuration together with Skills, so your AI system loads and uses them automatically once installed.
Manual upload: Download the Skills files from the repository and upload them to your AI tool by following your provider's documentation (see the table below).