Model Context Protocol

Connect Pluggy to your AI agent

Plug Claude, Cursor, ChatGPT — any AI agent — straight into Pluggy's live documentation and API reference. Grounded answers with citations, always current. No API key — use it anonymously for docs, or sign in with your Pluggy dashboard to reach your dev portal.

https://v2.docs.pluggy.ai/api/mcp

Connect your agent Read the docs

Answers with citations

Your agent answers from the live docs — guides, changelog, recipes, verified Q&A — and links its sources, instead of guessing. No login needed.

Always-current API reference

Every endpoint from the OpenAPI spec — methods, params, schemas — up to date the moment the docs change. No re-teaching your agent.

Debug a failing item

Sign in and ask why an item is failing — the agent reads its status and executions and tells you the cause and the fix (reconnect, consent, MFA, incident).

Proactive monitoring

Query your dashboard stats in plain language: connection health, errors by institution, sync lag, usage — over any range, comparing periods, spotting anomalies.

Connect it to your tool

Pick your agent — the server URL is always https://v2.docs.pluggy.ai/api/mcp.

Claude Code

claude mcp add --transport http pluggy-docs https://v2.docs.pluggy.ai/api/mcp

Add -s user to make it available in every project.

What you can ask

Pluggy MCPconnected

Why is item abc-123 failing, and what should the user do?

debug_item

It's in LOGIN_ERROR — the bank credentials expired (last successful sync 5 days ago). Ask the user to reconnect via Update Mode; no code change needed on your side.

How's my connection health this month vs last month?

get_stats ×2

Success rate is 94%, up from 91% last month. Most of the remaining errors are Itaú MFA timeouts — concentrated in the last week.

Talk to your stats — build your own dashboard

The dashboard shows fixed charts. Over MCP, every one of its metrics is a tool your agent can query on demand — so instead of hunting through screens, you just ask. Slice by connector, filter by error, pick any date range, compare periods. Each customer composes the exact view they need, in the client they already use: a Monday-morning health check, a Slack digest of the week's errors, an alert when a bank's failure rate spikes. It's your data, your questions, your dashboard — assembled by conversation, not by us.

Your connection healthlast 30 days

Success rate

94%+3

Errors by institution

Itaú

Bradesco

Nubank

Median sync lag

2.4hfaster

Success rate by institution, last 30 daysSync lag across my connectors right nowUsage vs the same week last monthAlert me when error rate > 5%

Build on Pluggy with your agent

One URL for everyone: anonymous for the docs, or sign in with your Pluggy dashboard account to also reach your dev portal — your apps, items, and usage, scoped to your team.

MCP documentation API Reference