ramp-mcp: A Ramp MCP server

What is ramp-mcp?

A Model Context Protocol server for retrieving, analyzing, and running tasks on Ramp data using the Ramp Developer API. It implements a simple ETL pipeline with an ephemeral in-memory SQLite database, allowing an LLM to query and analyze data while working around token and input size limitations. It’s built for developers integrating Ramp with AI assistants like Claude.

How to use ramp-mcp?

Clone the repository, install uv, and configure a Ramp Developer API client with the required scopes. Run the server from the CLI by setting RAMP_CLIENT_ID, RAMP_CLIENT_SECRET, and optionally RAMP_ENV (defaults to demo), then executing uv run ramp-mcp -s <COMMA-SEPARATED-SCOPES>. For Claude Desktop, add a configuration entry to claude_desktop_config.json with the appropriate command, args, and env values.

Key features of ramp-mcp

• Database tools: process_data, execute_query, clear_table for ephemeral SQLite.
• Fetch tools: get_ramp_categories, get_currencies for direct data retrieval.
• Load tools: loads 11 data types (transactions, reimbursements, bills, etc.) via scoped API endpoints.
• Uses OAuth 2.0 client credentials grant for authentication.
• Operates against a demo environment by default; switch to production with RAMP_ENV=prd.
• In-memory data store resets after each session.

Use cases of ramp-mcp

• Load and query Ramp transaction data to analyze spending patterns.
• Fetch and categorize business expenses for reporting or auditing.
• Examine vendor, department, or location data alongside spend limits and programs.
• Combine multiple scoped data sources in one ephemeral database for cross-referencing.
• Integrate Ramp data into an AI assistant for natural-language financial queries.

FAQ from ramp-mcp

What environment does ramp-mcp use by default?

The server runs against the Ramp demo environment unless RAMP_ENV=prd is set.

What are the runtime requirements?

You need uv for package management and execution, plus a Ramp Developer API client ID and secret with appropriate scopes enabled.

Where does the data live during a session?

All loaded data is stored in an ephemeral SQLite database in memory. It is deleted when the session ends.

Are there any limitations with large datasets?

Yes. Large datasets may not be processable due to API rate limits or your MCP client’s token/timeout constraints. Prompting the LLM to keep responses concise is recommended.

How is authentication handled?

Authentication uses OAuth 2.0 client credentials. The server requires RAMP_CLIENT_ID and RAMP_CLIENT_SECRET environment variables, and the requested API scopes must match those granted to the client.