Ramp MCP overview

Overview

Ramp MCP lets you securely connect Ramp with AI tools like Claude, ChatGPT, and Notion. Whether it is analyzing spend trends, auditing transactions, completing expenses, or managing approvals, Ramp allows users to manage their finances from wherever they are already working.

Your Ramp permissions come with you. The Ramp connection respects each user's role and access. The AI tool can only see data and take actions that the signed-in user can already perform in Ramp, and every write is recorded in the Ramp Audit Log. Admins can further restrict access by role, department, or user — see Managing employee access.

Example prompts you can try:

If you're a finance admin or bookkeeper:

If you're an employee:

Set up Ramp MCP

Choose your AI tool below. You need a Ramp account and access to your chosen AI platform. Want to try it with sample data first? See the FAQ below.

Claude

One-click installation from Claude's MCP directory. Sign in to your Ramp account when prompted.

ChatGPT

One-click installation from the ChatGPT app directory. Requires a ChatGPT Plus or Pro subscription.

Notion

Notion AI can connect to Ramp MCP as an external data source. You can install Ramp from Notion's connector marketplace, or use a custom configuration in Notion's MCP connections settings with the server URL https://mcp.ramp.com/mcp. See Notion's documentation for configuring external MCP servers.

Coding agents

For coding agents like Claude Code, Codex, and Cursor, we recommend using the Ramp CLI.

Other MCP-compatible tools

Any MCP-compatible AI client can connect using the server URL https://mcp.ramp.com/mcp. Check your tool's MCP documentation for the specific configuration format.

Your first prompt

Once you're connected, the easiest way to confirm everything is working is to just ask something. Try one of these to start:

From there, experiment. Natural language works — you don't need to know the "right" phrasing. If something doesn't work the first time, try rewording it or asking the assistant what it can do with Ramp. For more ideas, see What you can do with Ramp below.

Using a coding agent? Confirm Ramp appears in your connected MCP servers list before you start prompting.

Managing employee access

Admins can control which employees have access to Ramp MCP:

  1. Go to Company > Integrations > Ramp MCP .
  2. Select Manage Access .
  3. Set restrictions based on roles, departments, or specific users.

What you can do with Ramp

For Finance Admins and Bookkeepers

Close the books faster. Find missing receipts and memos, apply GL coding to transactions and reimbursements, tag tracking categories, leave comments, and ping team members when something needs attention.

Stay on top of spend. Pull top vendors, flag anomalies, export full spend data, and check Ramp Banking and Ramp Checking Account balances for a cash snapshot.

Review and approve. See what's pending your review and take action on transactions, reimbursements, and unified requests (purchase orders and fund requests).

Note: Bill approvals are not currently available through the AI connection. Use Ramp directly to approve bills.

Load and query your data. Load spending limits, Spend Programs, vendors, users, departments, entities, purchase orders, and full spend exports into a queryable workspace — then ask questions across all of it.

For Employees

Finish your expenses. Add memos, split categories, assign to a trip, tag attendees, and submit or resubmit reimbursements — all in chat. Upload receipt files directly in Ramp because MCP connections do not currently support passing file uploads, such as receipt images or PDFs, to Ramp MCP.

Manage your cards. See what you have, lock or unlock a card, activate a card by the last four digits, and generate agent card credentials for one-time payments.

Check balances and limits. See your funds, budgets, and remaining limits before you spend.

Plan and reconcile trips. Create a trip, view your flight and hotel bookings, and connect transactions to the right trip.

Get answers about Ramp and policy. Ask how something works, whether a spend is within policy, or pull an answer from the Ramp Help Center.

Full capability reference

The detailed list of what's supported today. Tools are added regularly — if something you'd expect isn't here, let us know.

Transactions

Reimbursements

Bills

Cards and funds

Spending limits and programs

Purchase orders and requests

Trips

Ramp Banking and accounts

Organization and people

Vendors and categories

Accounting

Comments and communication

Policy and help

Other

Not seeing a tool you expected? Try disconnecting and reconnecting Ramp in your AI tool — new tools are added regularly, and your session may need a fresh connection to pick them up. You can also submit feedback directly from any connected AI tool by asking it to "submit feedback to Ramp about [your feedback here]."

Troubleshooting

Authentication fails or connection drops

Try the fix for your specific tool first:

In every case, make sure you are signing in to the correct Ramp account (production vs. sandbox).

ETL operation limit reached

This error means too many data tables are loaded at once. Ask the AI assistant to clear unused tables (for example, "clear any tables I'm not using"), then retry your request. If the error continues, wait a few minutes for current data operations to complete.

Query returned more than 100 rows

The AI assistant will automatically add more specific filters and retry the query. You can also narrow your request -- for example, add a date range or filter by vendor.

Tools not appearing or not responding

For coding agents only

The following steps apply only to coding-agent setups. If you are using Claude, ChatGPT, or Notion, skip this section.

Permission errors

If a tool returns a permissions error, your Ramp role may not have access to that tool. Admin-only tools (such as company-wide data loading and large-dataset queries) require Admin or Business Owner permissions.

If you believe you should have access, ask your Ramp Admin to check your access settings in Company > Integrations > Ramp MCP > Manage Access.

For developers and IT Admins

What is the difference between Ramp MCP and the Ramp Developer API?

Ramp MCP is designed for use with AI assistants — it lets you interact with Ramp data using natural language through tools like Claude and ChatGPT. The Ramp Developer API is a REST API for building custom integrations and applications programmatically. In many cases, Ramp MCP already supports actions (like editing transactions or approving reimbursements) that would otherwise require building a custom app against the API. If you are not sure which to use, start with Ramp MCP.

Where can I find the Ramp CLI?

The Ramp CLI is a separate terminal tool for developers. Install it by running curl -fsSL https://agents.ramp.com/install.sh | sh in your terminal. Documentation is available at docs.ramp.com.

How do I get my redirect URI allowlisted for a custom MCP client?

If you are building a custom MCP client that connects to Ramp, your redirect URI must be allowlisted. To request this, contact our developer support team with the exact redirect URI(s) you need added. Note the following requirements:

Can I connect Ramp MCP through a third-party MCP gateway or proxy?

Yes, but your gateway's redirect URI must be added to the allowlist by Ramp first. If you are using a service like Glean, MintMCP, GoSearch, or another MCP gateway, contact our developer support team with the redirect URI your gateway uses. If your organization builds and hosts MCP proxies, see the redirect URI FAQ above for requirements.

Can I use Ramp MCP with Microsoft Copilot for 365?

Microsoft Copilot for 365 does not natively support remote MCP servers in the same way that Claude and ChatGPT do. To connect Ramp data to Copilot, you can use Copilot Studio's MCP server integration to add Ramp MCP as a tool for a Copilot agent. Alternatively, you can connect to the Ramp Developer API using Copilot Studio's custom API connector or Power Automate.

Frequently asked questions

I only see a few tools (like "get current user" and "ask Help Center"). Where are the rest?

This usually means your Ramp account does not have Admin or Business Owner permissions. Most tools (company-wide transactions, bills, reimbursements, spend data) require admin-level access. Ask your Ramp Admin to check your role, or verify your access in Company > Integrations > Ramp MCP > Manage Access.

Can the AI assistant approve things on my behalf?

Yes. If you are an approver, the Ramp connection can approve or reject transactions, reimbursements, and unified requests (including purchase orders and fund requests) on your behalf. Bill approvals are not currently supported through the connection — use Ramp directly for those. All actions respect your existing Ramp permissions and are recorded in the Ramp Audit Log.

Why doesn't Ramp MCP support receipt uploads?

Ramp MCP does not currently expose receipt upload as an MCP tool because current MCP connections do not provide a standard file-upload path for Ramp MCP to receive user-selected files from an AI chat. If your AI tool lets you attach a file to the conversation, that attachment is not automatically passed to Ramp MCP. This is a limitation of MCP file input support, not a limitation of Ramp's receipt upload feature. To upload receipts, use Ramp directly; see Submitting receipts and expenses. Developers building a custom integration can also use the Ramp Developer API receipt endpoint.

Can I try Ramp MCP with a sandbox or demo account?

Yes. You can use a sandbox Ramp account with the same setup steps above — just sign in with your sandbox credentials.

To explore with sample data and no Ramp account at all, use the demo server https://demo-mcp.ramp.com/mcp:

Which AI tools are compatible?

Any tool that supports the Model Context Protocol (MCP) can connect to Ramp. The setup flows above cover Claude, ChatGPT, Notion, and coding agents like Claude Code, Codex, and Cursor via the Ramp CLI. For the full setup instructions, see Set up Ramp MCP.

How do I get help?

For questions or issues, contact our developer support team. You can also submit feedback directly from your AI tool by asking it to "submit feedback to Ramp."