Plaid Data Partner MCP

The Plaid Data Partner MCP lets you use an AI agent to review production FDX request activity, troubleshoot individual calls, monitor integration health, and run and inspect validation for Core Exchange integrations.

Use an MCP client that supports remote servers. The server URL is:

https://data-partner.mcp.plaid.com/mcp

If your client asks for a transport, choose Streamable HTTP.

If no team appears after you sign in, confirm that your Dashboard account is verified and belongs to the expected Data Partner team.

claude mcp add --transport http --scope user plaid-dp-mcp https://data-partner.mcp.plaid.com/mcp

Run /mcp, select plaid-dp-mcp, and complete the browser sign-in. If you've already added the server as a connector in Claude, Claude Code loads it automatically when you sign in with the same account.

Go to Customize → Connectors → Add custom connector and enter the server URL.

On Team and Enterprise plans, an Owner must first add the URL under Organization settings → Connectors. See Claude's custom connector documentation.

In the app or IDE, open Settings → MCP servers, choose Streamable HTTP, and add the server URL. Or run:

codex mcp add plaid-dp-mcp --url https://data-partner.mcp.plaid.com/mcp
codex mcp login plaid-dp-mcp

Restart the app or IDE if prompted, then complete the browser sign-in.

Codex clients on the same computer share MCP configuration, so you only need to add the server once.

1. Go to Settings → Security and login → Enable Developer mode.
2. Go to chatgpt.com/plugins, select the + button, and create a developer-mode app using the server URL.

If you don't see Developer mode under Settings → Security and login, your plan doesn't support adding MCP servers or your workspace admin hasn't enabled it.

All tools require that the logged-in user have the corresponding permission in the Dashboard. If you encounter an authentication error, check that you are logged in as the correct user and that you have the required permissions.

• Search production FDX requests from the past 14 days by end user, endpoint, HTTP status, error state, or time range.
• Look up any request by ID to see its endpoint, status, latency, timestamp, and error details.

Response bodies are available only for failed requests.

Platform teams can search all their institutions or specific ones, and find institutions by name, ID, or integration status.

Example prompts:

• Look up request 0asd1a92nc, explain what happened, and suggest the next check.
• User <user_id> says balances are missing. Review their /accounts requests from the past 24 hours and explain the likely cause.
• For Gingham Bank, review failed /accounts requests from the past 24 hours and explain the likely cause.

• Start validation with an OAuth access token your institution issued for a test user.
• Single-institution teams can validate against production, staging, or development; platform teams can validate in production only.
• Review the results of recent validations, including scheduled ones, filtered by product or outcome.

Example prompt:

• Run extraction validation in production with <oauth_access_token>, then summarize failures and next steps.

• View successful and failed extractions from the past 14 days
• View the error rate over time
• View the most frequent extraction errors and whether each is new or increasing

Platform results can be narrowed to specific institutions and show which institutions produced each error.

Example prompt:

• Is my integration healthy? If error rates look elevated, show me what's failing.