
Plaid Sandbox MCP (Python)
OfficialPlaid's local MCP server to generate mock financial data and exercise sandbox APIs.
Add to your client
Copy the config for your MCP client and paste it into its config file.
uvx mcp-server-plaid --client-id <your-plaid-client-id> --secret <your-plaid-sandbox-secret>Paste into ~/Library/Application Support/Claude/claude_desktop_config.json
{
"mcpServers": {
"plaid-sandbox-mcp-python": {
"command": "uvx",
"args": [
"mcp-server-plaid",
"--client-id",
"<your-plaid-client-id>",
"--secret",
"<your-plaid-sandbox-secret>"
]
}
}
}Requires `uv` (the Python package runner). Install it from https://docs.astral.sh/uv/ if `uvx` is not found.
Step-by-step guides: Add to Claude Desktop · Add to Cursor · Add to Windsurf
Before you start
- A Plaid developer account with Sandbox API credentials (`client_id` and `secret`) from the Plaid Dashboard under Developers → Keys
- `uv`/`uvx` installed (the Python uv toolchain) to run `mcp-server-plaid`
- An MCP-compatible client such as Claude Desktop or Cursor
About Plaid Sandbox MCP (Python)
Plaid Sandbox MCP is the official local MCP server in Plaid's ai-coding-toolkit repository. It gives AI coding assistants a focused set of tools for building and testing Plaid integrations against Plaid's Sandbox environment, without touching real bank accounts or production data.
Rather than wrapping every Plaid product endpoint, it concentrates on the parts of the workflow that benefit most from an AI agent: searching Plaid's documentation, minting Sandbox access tokens, generating realistic mock financial data, and firing simulated webhooks so you can exercise your handler logic end to end.
It runs locally over stdio (typically launched with uvx mcp-server-plaid) and authenticates with your Sandbox client_id and secret. It is aimed at developers prototyping Plaid features inside Claude Desktop, Cursor, or any other MCP-compatible client.
Tools & capabilities (4)
search_documentationSearches Plaid's documentation for the right product, API, and endpoint information.
get_sandbox_access_tokenCreates a Sandbox Item and returns a working access token and item ID for testing against mocked data.
get_mock_data_promptReturns a prompt for generating customized, realistic mock financial test data.
simulate_webhookTriggers simulated webhook events so you can test your application's webhook-handling logic.
When to use it
- You are wiring up a new Plaid integration and want an agent that can pull the correct docs and endpoints on demand
- You need a Sandbox access token and item to test transactions, balances, or auth flows without real bank data
- You want realistic mock financial data to seed and demo your app
- You are building webhook handlers and need to replay simulated Plaid webhook events to verify they fire correctly
Quick setup
- 1Sign in to the Plaid Dashboard and copy your Sandbox `client_id` and `secret` from Developers → Keys
- 2Ensure `uv`/`uvx` is installed on your machine
- 3Add the server to your MCP client config with command `uvx` and args `mcp-server-plaid --client-id YOUR_CLIENT_ID --secret YOUR_SECRET`
- 4Restart the MCP client and confirm the four Plaid tools are listed
- 5Ask the agent to fetch a Sandbox access token or search the docs to verify it works
Security notes
Intended for Plaid sandbox use; supply sandbox credentials rather than production keys. Avoid pointing it at live Plaid secrets since it is a development/testing tool.
Plaid Sandbox MCP (Python) FAQ
Does this work against real bank accounts or production?
No. It is scoped to Plaid's Sandbox environment and uses your Sandbox credentials, so it generates and operates on mock financial data only — never real accounts.
Where do I get the client_id and secret?
From the Plaid Dashboard under Developers → Keys. Use the Sandbox credentials shown there.
How do I run it — npm or Python?
It is a Python package run via `uvx mcp-server-plaid`. You need the `uv` toolchain installed; there is no npm package.
Is it an official Plaid project?
Yes. It lives in Plaid's own `plaid/ai-coding-toolkit` repository and is MIT-licensed.
What can it actually do — does it cover every Plaid endpoint?
No. It intentionally exposes four tools focused on the integration workflow: documentation search, Sandbox token creation, mock-data prompts, and webhook simulation.
Alternatives to Plaid Sandbox MCP (Python)
Compare all alternatives →Official Stripe server for payments, customers, subscriptions, invoices, and billing via natural language.
Coinbase's AgentKit MCP extension for onchain actions: transfers, swaps, and contract calls.
Alpaca's official trading server for stocks, ETFs, options, and crypto in plain English.
Compare Plaid Sandbox MCP (Python) with: