generated: '2026-09-23'
method: searched
status: published
source:
  - https://plaid.com/docs/resources/mcp/index.html.md
  - https://github.com/plaid/ai-coding-toolkit/blob/main/sandbox/README.md
  - https://pypi.org/project/mcp-server-plaid/
description: >-
  Plaid publishes TWO official MCP servers. (1) The Dashboard MCP server is a
  remote, hosted MCP endpoint at https://api.dashboard.plaid.com/mcp/ that
  exposes production diagnostics and Link analytics tools; it is called via
  Streamable HTTP with an OAuth Bearer token (client_credentials, scope
  mcp:dashboard, obtained from /oauth/token). (2) The Plaid Sandbox MCP server
  (mcp-server-plaid, part of the Plaid AI Coding Toolkit) is a local stdio
  server intended to accelerate integration development with mock data,
  documentation search, sandbox tokens and webhook simulation. Both are
  official Plaid distributions; tools listed below are the exact tool names
  advertised in Plaid's own docs / README (2026-09-23).
servers:
  - name: plaid-dashboard-mcp
    role: dashboard-diagnostics-and-analytics
    transport: http-streamable
    url: https://api.dashboard.plaid.com/mcp/
    published: true
    auth:
      scheme: oauth2
      grant_type: client_credentials
      scope: mcp:dashboard
      token_endpoint: https://production.plaid.com/oauth/token
      token_ttl: 15m
    docs: https://plaid.com/docs/resources/mcp/
    tools:
      - name: plaid_debug_item
        description: Diagnose a Plaid Item using related metadata.
      - name: plaid_get_link_analytics
        description: Retrieve Link funnel, conversion, and error metrics.
      - name: plaid_get_tools_introduction
        description: Get usage guidance for Dashboard MCP tools.
      - name: plaid_get_usages
        description: Retrieve product usage metrics and API request volumes.
      - name: plaid_list_teams
        description: List teams available to the OAuth token.
  - name: plaid-sandbox-mcp
    role: sandbox-integration-development
    transport: stdio
    install: uvx mcp-server-plaid --client-id YOUR_PLAID_CLIENT_ID --secret YOUR_PLAID_SECRET
    package: https://github.com/plaid/ai-coding-toolkit/tree/main/sandbox
    registry: https://pypi.org/project/mcp-server-plaid/
    published: true
    auth:
      scheme: sandbox-credentials
      params: [PLAID_CLIENT_ID, PLAID_SECRET]
      note: requires Plaid Sandbox client_id + secret from the developer Dashboard
    docs: https://github.com/plaid/ai-coding-toolkit/blob/main/sandbox/README.md
    tools:
      - name: search_documentation
        description: Search Plaid documentation for relevant information about products or API endpoints.
      - name: get_mock_data_prompt
        description: Return a prompt to generate customized mock financial data for testing.
      - name: get_sandbox_access_token
        description: Obtain a working access_token + item_id for the Plaid Sandbox environment.
      - name: simulate_webhook
        description: Simulate a Plaid webhook event in the Sandbox environment for handler testing.
deployment:
  mode: both
  endpoint: https://api.dashboard.plaid.com/mcp/
  install: uvx mcp-server-plaid --client-id YOUR_PLAID_CLIENT_ID --secret YOUR_PLAID_SECRET
  package: https://github.com/plaid/ai-coding-toolkit/tree/main/sandbox
  auth: oauth
  verified: probed
  probes:
    - url: https://api.dashboard.plaid.com/mcp/
      method: HEAD
      status: 411
      note: expects a POST body per MCP Streamable HTTP; live endpoint confirmed
    - url: https://api.dashboard.plaid.com/mcp
      method: HEAD
      status: 307
      note: redirects to /mcp/
    - url: https://pypi.org/project/mcp-server-plaid/
      method: HEAD
      status: 200
    - url: https://github.com/plaid/ai-coding-toolkit
      method: HEAD
      status: 200
  checked: '2026-09-23'
notes: >-
  The Dashboard MCP server is under active development; Plaid documents that
  breaking changes may occur and support is limited. Production Plaid access
  with at least one approved product is required to use the Dashboard MCP
  server. Neither MCP server exposes the full Plaid public API surface: the
  Dashboard server is analytics/diagnostics scoped, and the Sandbox server is
  developer-tooling scoped. See mcp/plaid-tool-crosswalk.yml for how the
  advertised MCP tools map onto Plaid's OpenAPI operations.