Ramp
About
Interact with [Ramp](https://ramp.com)'s Developer API to run analysis on your spend and gain insights leveraging LLMs
Details
- License
- MIT
Explore
- 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.
Run the MCP server from your CLI with:
RAMP_CLIENT_ID=... RAMP_CLIENT_SECRET=... RAMP_ENV=<demo|prd> uv run ramp-mcp -s <COMMA-SEPARATED-SCOPES>
1. Create a new client from the Ramp developer page (Profile on top right > Developer > Create app)
2. Grant the scopes you wish (based on tools) to the client and enable client credentials (Click on App > Grant Types / Scopes)
3. Include the client ID and secret in the config file as well as the scopes you wish to use
1. Clone this Github repo via git clone [email protected]:ramp/ramp-mcp.git or equivalent
2. Install uv
Add this to your claude_desktop_config.json:
{
"mcpServers": {
"ramp-mcp": {
"command": "uv",
"args": [
"--directory",
"/<ABSOLUTE-PATH-TO>/ramp-mcp", // make sure to update this path
"run",
"ramp-mcp",
"-s",
"transactions:read,reimbursements:read"
],
"env": {
"RAMP_CLIENT_ID": "<CLIENT_ID>",
"RAMP_CLIENT_SECRET": "<CLIENT_SECRET>",
"RAMP_ENV": "<demo|qa|prd>"
}
}
}
}
If this file doesn't exist yet, create one in /<ABSOLUTE-PATH-TO>/Library/Application Support/Claude/
load_transactions
transactions:read
load_reimbursements
reimbursements:read
load_bills
bills:read
load_locations
locations:read
load_departments
departments:read
load_bank_accounts
bank_accounts:read
load_vendors
vendors:read
load_vendor_bank_accounts
vendors:read
load_entities
entities:read
load_spend_limits
limits:read
load_spend_programs
spend_programs:read
load_users
users:read
Tools that can be used to setup, process, query, and delete an ephemeral database in memory.
1. process_data
2. execute_query
3. clear_table
Tools that can be used to fetch data directly
1. get_ramp_categories
2. get_currencies
Loads data to server which the client can fetch. Based on the tools you wish to use, ensure to enable those scopes on your
Ramp client and include the scopes when starting the server as a CLI argument.
| Tool | Scope |
| ------------------------- | ------------------- |
| load_transactions | transactions:read |
| load_reimbursements | reimbursements:read |
| load_bills | bills:read |
| load_locations | locations:read |
| load_departments | departments:read |
| load_bank_accounts | bank_accounts:read |
| load_vendors | vendors:read |
| load_vendor_bank_accounts | vendors:read |
| load_entities | entities:read |
| load_spend_limits | limits:read |
| load_spend_programs | spend_programs:read |
| load_users | users:read |
For large datasets, it is recommended to explicitly prompt Claude not to use REPL and to keep responses concise to avoid timeout or excessive token usage.
Overview
A Model Context Protocol server for retrieving and analyzing data or running tasks for Ramp using Developer API. In order to get around token and input size limitations, this server implements a simple ETL pipeline + ephemeral sqlite database in memory for analysis by an LLM. All requests are made to demo by default, but can be changed by setting RAMP_ENV=prd. Large datasets may not be processable due to API and/or your MCP client limitations.
Tools
Database tools
Tools that can be used to setup, process, query, and delete an ephemeral database in memory.
1. process_data
2. execute_query
3. clear_table
Fetch tools
Tools that can be used to fetch data directly
1. get_ramp_categories
2. get_currencies
Load tools
Loads data to server which the client can fetch. Based on the tools you wish to use, ensure to enable those scopes on your
Ramp client and include the scopes when starting the server as a CLI argument.
| Tool | Scope |
| ------------------------- | ------------------- |
| load_transactions | transactions:read |
| load_reimbursements | reimbursements:read |
| load_bills | bills:read |
| load_locations | locations:read |
| load_departments | departments:read |
| load_bank_accounts | bank_accounts:read |
| load_vendors | vendors:read |
| load_vendor_bank_accounts | vendors:read |
| load_entities | entities:read |
| load_spend_limits | limits:read |
| load_spend_programs | spend_programs:read |
| load_users | users:read |
For large datasets, it is recommended to explicitly prompt Claude not to use REPL and to keep responses concise to avoid timeout or excessive token usage.
Setup
Ramp Setup
1. Create a new client from the Ramp developer page (Profile on top right > Developer > Create app)
2. Grant the scopes you wish (based on tools) to the client and enable client credentials (Click on App > Grant Types / Scopes)
3. Include the client ID and secret in the config file as well as the scopes you wish to use
Local Setup
1. Clone this Github repo via git clone [email protected]:ramp/ramp-mcp.git or equivalent
2. Install uv
Usage
Run the MCP server from your CLI with:
RAMP_CLIENT_ID=... RAMP_CLIENT_SECRET=... RAMP_ENV=<demo|prd> uv run ramp-mcp -s <COMMA-SEPARATED-SCOPES>
Configuration
Usage with Claude Desktop
Add this to your claude_desktop_config.json:
{
"mcpServers": {
"ramp-mcp": {
"command": "uv",
"args": [
"--directory",
"/<ABSOLUTE-PATH-TO>/ramp-mcp", // make sure to update this path
"run",
"ramp-mcp",
"-s",
"transactions:read,reimbursements:read"
],
"env": {
"RAMP_CLIENT_ID": "<CLIENT_ID>",
"RAMP_CLIENT_SECRET": "<CLIENT_SECRET>",
"RAMP_ENV": "<demo|qa|prd>"
}
}
}
}
If this file doesn't exist yet, create one in /<ABSOLUTE-PATH-TO>/Library/Application Support/Claude/
License
Copyright (c) 2025, Ramp Business Corporation
All rights reserved.
This source code is licensed under the MIT License found in the LICENSE file in the root directory of this source tree.
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.


