Marqeta DiVA API MCP Server

A Model Context Protocol (MCP) server that provides programmatic access to the Marqeta DiVA (Data insights, Visualization, and Analytics) API. This server enables AI assistants and applications to retrieve aggregated production data from the Marqeta platform for reporting, analytics, and data-driven business decisions.

Note: This is an unofficial community project and is not officially supported by Marqeta.

Features

Core Features (Always Available)

• Transaction Data: Access authorizations, settlements, clearings, declines, and loads
• Financial Data: Retrieve program balances, settlement balances, and activity balances
• Card & User Data: Get card and user details with flexible filtering
• Chargeback Data: Access chargeback status and detailed information
• Metadata Tools: Discover available views and retrieve schema definitions
• Export Tools: Export data to JSON or CSV files
• Rate Limiting: Built-in rate limiting to comply with API limits (300 requests per 5 minutes)
• Error Handling: Comprehensive error handling with meaningful messages
• Flexible Querying: Support for filtering, sorting, field selection, date ranges, and more

Optional RAG Features (Requires [rag] extras)

• Local Storage: Store complete transaction data in SQLite (bypasses MCP token limits)
• Semantic Search: Natural language queries on transaction data using AI embeddings
• Vector Store: ChromaDB integration for similarity-based transaction search
• Offline Analysis: Query local data without API calls or token limits

Prerequisites

• Python 3.10 or higher
• uv package manager (for running with uvx)
• Marqeta DiVA API credentials (Application Token, Access Token, and Program Name)

Installation

Option 1: Run with uvx (Recommended)

No installation needed! uvx will automatically handle dependencies when you run the server.

For basic features only:

uvx marqeta-diva-mcp

For RAG features (local storage + semantic search):

uvx --with marqeta-diva-mcp[rag] marqeta-diva-mcp

Option 2: Traditional Installation

Basic installation (core features only):

pip install marqeta-diva-mcp

With RAG features (recommended for advanced analytics):

pip install marqeta-diva-mcp[rag]

From source:

cd marqeta-diva-mcp
pip install -e .              # Basic features
pip install -e ".[rag]"       # With RAG features

Configuration

1. Copy the example environment file:

cp .env.example .env

2. Edit .env and add your Marqeta DiVA API credentials:

# Required: Marqeta DiVA API credentials
MARQETA_APP_TOKEN=your_application_token_here
MARQETA_ACCESS_TOKEN=your_access_token_here
MARQETA_PROGRAM=your_program_name_here

# Optional: Enable local storage and RAG features
# Requires: pip install marqeta-diva-mcp[rag]
# ENABLE_LOCAL_STORAGE=true

How to obtain credentials:

• Contact your Marqeta representative, OR
• Generate via Marqeta Dashboard (Reports section)

Enabling RAG Features:

To use local storage, semantic search, and other RAG features:

1. Install RAG dependencies: pip install marqeta-diva-mcp[rag]
2. Set environment variable: ENABLE_LOCAL_STORAGE=true
3. Restart the MCP server

When enabled, you'll see this message in the logs:

[MCP Server] Local storage and RAG features ENABLED

When disabled (default):

[MCP Server] Local storage and RAG features DISABLED (set ENABLE_LOCAL_STORAGE=true to enable)

Usage

Running the Server Locally

With uvx (Recommended)

cd marqeta-diva-mcp
uvx marqeta-diva-mcp

With Python

cd marqeta-diva-mcp
python -m marqeta_diva_mcp.server

Adding to Claude Desktop

Add this configuration to your Claude Desktop config file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%/Claude/claude_desktop_config.json

Using uvx (Recommended)

Basic configuration (core features only):

{
  "mcpServers": {
    "marqeta-diva": {
      "command": "uvx",
      "args": ["marqeta-diva-mcp"],
      "env": {
        "MARQETA_APP_TOKEN": "your_application_token",
        "MARQETA_ACCESS_TOKEN": "your_access_token",
        "MARQETA_PROGRAM": "your_program_name"
      }
    }
  }
}

With RAG features (local storage + semantic search):

{
  "mcpServers": {
    "marqeta-diva": {
      "command": "uvx",
      "args": ["--with", "marqeta-diva-mcp[rag]", "marqeta-diva-mcp"],
      "env": {
        "MARQETA_APP_TOKEN": "your_application_token",
        "MARQETA_ACCESS_TOKEN": "your_access_token",
        "MARQETA_PROGRAM": "your_program_name",
        "ENABLE_LOCAL_STORAGE": "true"
      }
    }
  }
}

Using Python

Basic configuration (core features only):

{
  "mcpServers": {
    "marqeta-diva": {
      "command": "python",
      "args": ["-m", "marqeta_diva_mcp.server"],
      "cwd": "/path/to/marqeta-diva-mcp",
      "env": {
        "MARQETA_APP_TOKEN": "your_application_token",
        "MARQETA_ACCESS_TOKEN": "your_access_token",
        "MARQETA_PROGRAM": "your_program_name"
      }
    }
  }
}

With RAG features (requires pip install -e ".[rag]" first):

{
  "mcpServers": {
    "marqeta-diva": {
      "command": "python",
      "args": ["-m", "marqeta_diva_mcp.server"],
      "cwd": "/path/to/marqeta-diva-mcp",
      "env": {
        "MARQETA_APP_TOKEN": "your_application_token",
        "MARQETA_ACCESS_TOKEN": "your_access_token",
        "MARQETA_PROGRAM": "your_program_name",
        "ENABLE_LOCAL_STORAGE": "true"
      }
    }
  }
}

Platform Integrations

This MCP server can be integrated with various AI platforms and tools. We provide comprehensive guides for:

MCP-Compatible Platforms

• Claude Desktop (see configuration above) - Native MCP support
• Claude Code - CLI with MCP support
• Cline - VS Code extension with MCP support
• Other MCP clients - Any client supporting the MCP protocol

Non-MCP Platforms

• ChatGPT / OpenAI - Using direct Python client, REST wrapper, or export methods
• Jupyter Notebooks - Direct client usage with pandas
• Python Scripts - Standalone script integration
• Custom Applications - REST API wrappers, Slack/Discord bots, web dashboards

Integration Guides

📚 INTEGRATIONS.md - Comprehensive integration guide covering:

• Detailed setup instructions for each platform
• Configuration examples and code snippets
• Troubleshooting tips
• Best practices for security and performance
• Custom integration patterns

⚡ QUICK_INTEGRATION.md - Quick reference guide with:

• 2-minute Claude Desktop setup
• 2-minute Claude Code setup
• 1-minute Python/Jupyter setup
• Fast troubleshooting tips

Available Tools

Transaction Tools

get_authorizations

Get authorization transaction data with amounts, counts, acting users/cards, and merchant information.

Parameters:

• aggregation (string): detail, day, week, or month (default: detail)
• start_date (string): Start date in ISO format (YYYY-MM-DD or YYYY-MM-DDTHH:MM:SS)
• end_date (string): End date in ISO format
• fields (array): Specific fields to return
• filters (object): Additional filters (e.g., {"transaction_amount": ">100"})
• sort_by (string): Field to sort by (prefix with - for descending)
• count (integer): Maximum records to return (up to 10,000)
• program (string): Override default program name

Example:

{
  "aggregation": "day",
  "start_date": "2024-01-01",
  "end_date": "2024-01-31",
  "filters": {"transaction_amount": ">1000"},
  "sort_by": "-request_amount",
  "count": 100
}

get_settlements

Get settlement transaction data including status, post dates, purchase amounts, and network information.

Parameters: Same as get_authorizations

get_clearings

Get accounting-level line items for transaction lifecycle. Ideal for reconciliation.

Parameters: Same as get_authorizations

get_declines

Get declined transaction data with tokens, decline reasons, merchant info, and amounts.

Parameters: Same as get_authorizations

get_loads

Get load transaction data including amounts and transaction details.

Parameters: Same as get_authorizations

Financial Tools

get_program_balances

Get program-level balance data including beginning/ending bank balances and amounts to send/receive.

Parameters:

• start_date, end_date, fields, filters, sort_by, count, program

get_program_balances_settlement

Get settlement-based program balance data and fund transfers.

Parameters: Same as get_program_balances

get_activity_balances

Get cardholder-level balance data, expandable by network.

Parameters:

• All standard parameters plus:
• expand (string): Field to expand for more detail (e.g., network data)

Card & User Tools

get_cards

Get card detail data including user tokens, card state, active status, and UAI.

Parameters:

• fields, filters, sort_by, count, program

Example filters:

{
  "filters": {
    "state": "ACTIVE",
    "user_token": "abc123"
  }
}

get_users

Get user detail data including tokens, UAI, and number of physical/virtual cards.

Parameters: Same as get_cards

Chargeback Tools

get_chargebacks_status

Get chargeback status data including state, tokens, and provisional credit status.

Parameters:

• start_date, end_date, fields, filters, sort_by, count, program

get_chargebacks_detail

Get detailed chargeback information with transaction dates and types.

Parameters: Same as get_chargebacks_status

Reconciliation Tools

get_transaction_token ⭐ NEW in v0.3.0

CRITICAL for reconciliation - Map Core API transaction tokens to DiVA report transaction tokens.

Purpose: Links webhook transaction data to DiVA reporting data. Essential for financial reconciliation workflows.

Parameters:

• fields, filters, sort_by, count, program

Example:

{
  "filters": {
    "core_api_transaction_token": "xyz123"
  }
}

Monitoring Tools

get_card_counts 📊 NEW in v0.3.0

Get card count metrics aggregated over time. Track cards in circulation, active, suspended, etc.

Parameters:

• aggregation (string): day, week, or month (required, no detail level)
• fields, filters, sort_by, count, program

Example:

{
  "aggregation": "day",
  "count": 30
}

get_user_counts 📊 NEW in v0.3.0

Get user count metrics aggregated over time. Track user base growth and engagement.

Parameters:

• aggregation (string): day, week, or month (required, no detail level)
• fields, filters, sort_by, count, program

Example:

{
  "aggregation": "week",
  "filters": {
    "user_type": "BUSINESS"
  }
}

Network Analytics Tools

get_activity_balances_network_detail 🌐 NEW in v0.3.0

Get activity balance data broken out by card network (Visa, Mastercard, Maestro, Cirrus, etc.).

Purpose: Understand network-specific transaction volumes. Day aggregation only.

Parameters:

• fields, filters, sort_by, count, program
• expand (string): pin_purchases_net or sig_purchases_net (comma-delimited for multiple)

Example:

{
  "expand": "pin_purchases_net,sig_purchases_net",
  "count": 10
}

Metadata Tools

list_available_views

Get a list of all available DiVA API view endpoints with metadata.

Parameters: None

get_view_schema

Get the schema definition for any view endpoint with field names, types, and descriptions.

Parameters:

• view_name (string, required): Name of the view (e.g., authorizations, settlements, cards)
• aggregation (string): Aggregation level if applicable (default: detail)

Query Filtering

The DiVA API supports powerful filtering operators:

Operator	Description	Example
~	Like (wildcard)	{"company": "Mar~eta"}
..	Range	{"date": "2023-10-01..2023-10-03"}
<, <=	Less than	{"amount": "<=100"}
>, >=	Greater than	{"date": ">=2023-04-01"}
=	Equal/In list	{"amount": "0"} or {"country": "United States,Mexico"}
=!	Not equal/Not in	{"amount": "=!0"}

Example:

{
  "filters": {
    "transaction_amount": ">1000",
    "post_date": "2023-02-01..2023-02-28",
    "state": "COMPLETION"
  }
}

Rate Limits

• Maximum: 300 requests per 5-minute interval (≈1 per second)
• Enforcement: Built-in rate limiter automatically throttles requests
• Error Code: HTTP 429 if limit exceeded

Data Synchronization

Report data is synchronized 3 times daily. See Marqeta documentation for specific refresh timelines.

Error Handling

The server handles all common DiVA API errors:

Code	Description
400	Bad Request - Malformed query or filter
403	Forbidden - Unauthorized access to field, filter, or program
404	Not Found - Malformed URL or endpoint doesn't exist
429	Rate limit exceeded

Example Usage with Claude

Once configured in Claude Desktop, you can use natural language queries:

Example queries:

• "Get all authorization transactions from last week with amounts over $1000"
• "Show me the settlement data for January 2024"
• "List all active cards for user token abc123"
• "What are the available views in the DiVA API?"
• "Get the schema for the settlements view"
• "Show me chargeback status for the last 30 days"
• "Get program balances for February 2024"

API Documentation

For complete DiVA API documentation, visit: https://www.marqeta.com/docs/diva-api/introduction/

Troubleshooting

Missing Credentials Error

Error: Missing required environment variables: MARQETA_APP_TOKEN, MARQETA_ACCESS_TOKEN, MARQETA_PROGRAM

Solution: Ensure your .env file exists and contains all three required variables.

Authentication Error (403)

Error 403: Forbidden - Unauthorized access

Solution: Verify your Application Token and Access Token are correct. Check that you have access to the specified program.

Rate Limit Error (429)

Error 429: Rate limit exceeded - Maximum 300 requests per 5 minutes

Solution: The built-in rate limiter should prevent this, but if you see it, wait a few minutes before making more requests.

Development

Running Tests

pytest

Code Formatting

black src/
ruff check src/

Contributing

Contributions are welcome! Please feel free to submit a Pull Request. For major changes, please open an issue first to discuss what you would like to change.

Development Setup

# Clone the repository
git clone https://github.com/zvika-finally/marqeta-diva-mcp.git
cd marqeta-diva-mcp

# Install with development dependencies
pip install -e ".[dev,rag]"

# Run tests
python test_fixes_unit.py

# Format code
black src/
ruff check src/

License

This project is licensed under the MIT License - see the LICENSE file for details.

Author

Zvika Badalov - zvika.badalov@finally.com

Acknowledgments

• Built with Model Context Protocol (MCP)
• Powered by Marqeta's DiVA API

Support

• Issues: GitHub Issues
• Marqeta API Questions: Contact your Marqeta representative or refer to the official Marqeta documentation

Disclaimer

This is an unofficial community project and is not officially endorsed or supported by Marqeta, Inc. Use at your own risk.