MCP Server

Connect AI agents directly to your PIM

Sigma PIM ships an MCP (Model Context Protocol) server that lets Claude, Cursor, or any MCP-compatible client browse products, propose changes, and interact with reviewers — without writing integration code.

How it works

The MCP server is a stdio subprocess that wraps the Sigma PIM REST API. MCP clients (like Claude Code or Cursor) launch it automatically and call tools to interact with your product data. All changes proposed by AI agents go through the Change Sets approval workflow, so humans always stay in control.

Protocol

Model Context Protocol (MCP) over stdio. No HTTP server, no ports to open.

Authentication

API key via SIGMA_API_KEY env var. The same keys from API Keys.

Requirements

Node.js 20+ and a running Sigma PIM API instance (local or deployed).

Safety

AI agents propose changes via Change Sets. Humans review and approve before anything is applied.

Setup

Add this to your project's .mcp.json file (or equivalent MCP client config):

{
  "mcpServers": {
    "sigma-pim": {
      "command": "node",
      "args": ["packages/mcp/dist/index.js"],
      "env": {
        "SIGMA_API_URL": "https://api.sigma-pim.com/api/v1",
        "SIGMA_API_KEY": "your-api-key-here"
      }
    }
  }
}

Then build the server:

pnpm --filter @sigma-pim/mcp build

Restart your MCP client (e.g. Claude Code) and the tools will be available immediately.

18 Available Tools

Understand the Data Model

list_schemas

List all product schemas — understand what product types exist.

get_schema

Get a schema with all attribute definitions (types, validation, scoping).

Browse & Search Products

list_records

List products with pagination. Optionally filter by schema.

get_record

Get a single product with all scoped attribute values.

search_records

Structured search (7 operators) or full-text search across products.

get_completeness

Check what required fields are missing on a product.

Propose Changes via Change Sets

This is the core value of the MCP integration. AI agents propose changes that go through human review before being applied to products.

create_change_set

Create a change set with proposed create/update/delete operations.

submit_change_set

Submit a draft for human review.

get_change_set

Check approval status and reviewer feedback.

list_change_sets

List change sets filtered by status or source type.

send_message

Message reviewers — explain reasoning, respond to revision requests.

get_messages

Read reviewer feedback and revision request messages.

Direct Edits & Organization

update_record_attribute

Directly update one attribute on a product (bypasses approval).

list_catalogs

List all product catalogs.

get_catalog_tree

Get the category hierarchy for a catalog.

Observe

get_change_set_stats

Dashboard stats — pending, approved, rejected counts.

get_record_audit

Audit history for a product — who changed what and when.

MCP Resources

The MCP server also exposes reference data as resources. MCP clients can read these to understand the PIM's configuration without making tool calls.

URI	Description
`sigma://dimensions`	All dimensions (Language, Market, Channel) with their segments. Needed to understand scoped attribute values.
`sigma://option-lists`	All option lists with allowed values. Needed when setting option-type attributes.
`sigma://workflows`	Workflow definitions with lifecycle states and transitions.

Example Workflow

A typical AI agent workflow using the MCP server:

// 1. Understand the data model
list_schemas()
  → "Apparel (12 attributes), Electronics (18 attributes)"

// 2. Get schema details to know what attributes exist
get_schema("schema-apparel")
  → { name, description, price, material, ... }

// 3. Find products that need work
search_records({ criteria: [{ field: "material", operator: "eq", value: "" }] })
  → 23 products missing material

// 4. Propose changes via a change set
create_change_set({
  title: "Fill missing material attributes",
  operations: [
    { type: "update", entityId: "ent-001", changes: [
      { attributeAlias: "material", newValue: "100% organic cotton" }
    ]},
    // ... more operations
  ]
})
  → Change set created: cs-abc123

// 5. Submit for human review
submit_change_set("cs-abc123")
  → Submitted for review

// 6. Check back later for feedback
get_messages("cs-abc123")
  → [reviewer] "Looks good, but use 'Organic Cotton' not '100% organic cotton'"

// 7. Respond and iterate
send_message("cs-abc123", "Updated — thanks for the feedback!")

Use Cases

Product Enrichment

AI agents fill in missing descriptions, materials, care instructions, and other attributes based on product names and images.

Translation

Translate scoped text attributes across all language dimensions. Propose translations as a change set for native speaker review.

Data Quality

Search for incomplete products, check completeness scores, and propose fixes. Use audit trails to track improvement over time.

Catalog Management

Browse catalog trees, identify uncategorized products, and suggest category assignments based on product attributes.

Previous Webhooks Next Errors