This repository provides a Model Context Protocol server for Ramp, enabling data retrieval, analysis, and task execution using Ramp's Developer API. It uses an ETL pipeline and in-memory SQLite database
This Model Context Protocol (MCP) server facilitates data retrieval and analysis for Ramp using the Developer API. It overcomes token and input size limitations by implementing an ETL pipeline with an in-memory SQLite database for LLM analysis. By default, requests target the demo environment, switch to production by setting RAMP_ENV=prd
. Note that large datasets might exceed API or client limitations.
The server provides tools for database management (setup, query, clear), data fetching (get_ramp_categories
, get_currencies
), and data loading. Loading tools require specific Ramp client scopes, as outlined in the table, such as transactions:read
for load_transactions
. For large datasets, advise Claude to avoid REPL and provide concise responses to prevent timeouts.
Create a Ramp developer client, grant necessary scopes, and enable client credentials.
Clone the repository using git clone [email protected]:ramp/ramp-mcp.git
and install uv
.
Run the server via CLI:
RAMP_CLIENT_ID=... RAMP_CLIENT_SECRET=... RAMP_ENV=<demo|prd> uv run ramp-mcp -s <COMMA-SEPARATED-SCOPES>
Configure claude_desktop_config.json
with the server command, arguments (including the absolute path to the ramp-mcp directory), and environment variables (client ID, secret, and environment).
Licensed under the MIT License.
ramp-public/ramp-mcp
March 20, 2025
March 28, 2025
Python