# Plaid

> Official Plaid MCP — developer sandbox tooling and dashboard diagnostics, not a live bank-data connector.

[Canonical HTML page](https://top-mcps.com/mcp/plaid) · [server.json](https://top-mcps.com/mcp/plaid.json) · [methodology](https://top-mcps.com/about/methodology)

## At a glance

- **Maintainer:** Plaid
- **Transport:** stdio, Streamable HTTP
- **Auth model:** OAuth 2.1
- **Required secrets:** None
- **Supported clients:** Claude, Cursor, VS Code, Zed, Codex, Any MCP-compatible client
- **License:** MIT
- **Language:** Python
- **Latest version:** latest
- **Last verified:** 2026-07-15
- **GitHub stars:** 28 (fetched 2026-07-17T12:42:33.113Z)
- **Score:** 78/100 (rubric 2026-04 — see https://top-mcps.com/about/methodology)
- **Source:** https://github.com/plaid/ai-coding-toolkit

## Security & scope

- **Access scope:** network
- **Sandbox:** AI Coding Toolkit operates on sandbox/mock data only. Dashboard MCP authenticates with an OAuth token scoped to mcp:dashboard (client-credentials grant, 15-minute expiry).
- **Gotchas:**
  - Do not expect live bank-account access — that is a common misread of "Plaid MCP". These servers are for building and diagnosing integrations, not aggregating account data.
  - Dashboard MCP tokens expire after 15 minutes and must be refreshed; the toolkit is Python, so it needs a Python runtime rather than npx.

## Quick answer

**What it does.** AI Coding Toolkit: mock-data generation, documentation search, sandbox access tokens, and webhook simulation. Dashboard MCP: plaid_debug_item, plaid_get_link_analytics, plaid_get_usages, plaid_list_teams, plus usage guidance.

**Best for:**
- Building a Plaid integration with AI assistance
- Generating sandbox/mock financial data for tests
- Searching Plaid documentation from an agent
- Debugging Items and Link conversion in production

**Not for:**
- Reading a live consumer's bank transactions or balances
- Personal-finance dashboards over real accounts
- Payment initiation or money movement

## Description

Plaid ships two official MCP servers, and both are for builders rather than end-user finance. The AI Coding Toolkit (local) generates mock financial data, searches Plaid docs, mints sandbox tokens, and simulates webhooks; the Dashboard MCP (remote) surfaces production diagnostics — debugging Items, Link conversion analytics, and API usage. Neither reads a live consumer's bank transactions or balances.

## Why it matters

Plaid is the default bank-connectivity API for fintech. These MCPs speed up building and operating a Plaid integration from inside an agent — but they are development and observability tools, not a way to pull a user's account balances.

## Key features

- Two official Plaid MCPs (dev toolkit + dashboard)
- Sandbox data generation and webhook simulation
- Docs search inside the agent
- Production diagnostics and Link analytics
- MIT-licensed toolkit

## FAQ

### Can the Plaid MCP read my bank transactions?

No. The official Plaid MCPs are developer and dashboard tools. The AI Coding Toolkit works only against sandbox/mock data, and the Dashboard MCP surfaces production diagnostics and analytics. Reading a live consumer's transactions requires the Plaid API directly, not these MCP servers.

### Which one should I use?

Use the AI Coding Toolkit while building — it generates sandbox data, searches docs, and simulates webhooks. Use the Dashboard MCP to operate a live integration — debug Items, review Link conversion, and check API usage.

### Is it official?

Yes — both are published by Plaid. The AI Coding Toolkit is open-source (MIT); the Dashboard MCP is a remote server hosted by Plaid at api.dashboard.plaid.com/mcp with OAuth client-credentials auth.

## Changelog

- **2026-07-15** — Refreshed install snippets and fact sheet; verified for 2026.
- **2025-06-01** — Initial directory listing.