Metricool MCP: Setup for Claude, ChatGPT and n8n, Tools and Limits

What Metricool's MCP actually is

MCP stands for Model Context Protocol, an open standard that lets AI tools and agents connect securely to platforms like Metricool. Metricool's MCP is its official server for that standard, and it lets an AI client read and act on your account: list your brands, pull metrics, check the calendar and create or change scheduled posts from a chat window ^[1]. The production address is ai.metricool.com/mcp. The code is open source, published on GitHub in the metricool/mcp-metricool repository and on PyPI. Metricool describes the server as a layer on top of its own API that removes the need to write code. That is the whole idea. You talk; the AI calls Metricool for you.

The clients Metricool documents are a long list: Claude Desktop, Claude.ai, Claude Code, ChatGPT, Cursor, Mistral Vibe, Make (the automation tool), Zapier, Smithery, TypingMind, n8n and Gemini CLI. This guide covers the three setups most readers ask about, Claude, ChatGPT and n8n, then what the server can and cannot touch, the scheduling rules that trip people up, what to do when it will not connect, and how it compares with the API and Metricool Studio. We read Metricool's help center, its MCP guides and its pricing page; we did not connect an account ourselves.

Every plan can use it, but your plan's limits come along

The short answer on cost is good news. The MCP works on every Metricool plan, including Free, and the help center says so on its setup, limits, ChatGPT and basics pages. The catch is that your plan limits still apply to whatever the AI queries or schedules ^[2]. A Free account connected through the MCP still stops at 20 posts a month and still sees only 30 days of analytics history. Free also covers just one brand and cannot connect LinkedIn or X, so the AI cannot reach those networks either. Asking a chatbot does not unlock anything the account itself lacks.

Two actions need more than Free. Sending a post for review through the MCP requires Advanced or higher, because content approval is an Advanced and Custom feature. And any token-based connection that goes through the Metricool API instead of a sign-in also requires Advanced. Advanced costs $67 a month billed monthly or $53 a month billed annually, for up to 15 brands. Starter, the first paid step, costs $25 a month billed monthly or $20 a month billed annually, for up to 5 brands. It adds LinkedIn but still has one user and no approvals. On plans bought on the web, X needs a paid plan plus the X add-on at $10 a month per connected X account (buyers from before July 2026 may keep the older $5 price); the app-only Starter Mobile 5 plan includes X in its monthly price.

Some older Metricool pages say otherwise. The April 2026 Terms still describe the AI connection as paid-only, a marketing page on MCPs says it needs Advanced or higher, and the Google Business Profile page lists the API and the MCP together under Advanced. The newer help pages and the September 2026 launch post say every plan, and the limits page states plainly that the MCP needs no paid plan. We follow the newer pages.

How we got here: from MetriLAB AI to open connectors

The history explains the mixed messages. In June 2025 Metricool launched MetriLAB AI, a Claude connection built on the Model Context Protocol, and it was available only to Advanced users who also had a Claude Pro subscription. Only a select set of actions worked at launch, and ChatGPT support was merely planned. By January 2026 Metricool's integrations page listed the MCP in the Claude Connectors Directory, with a one-click listing for ChatGPT. In September 2026 Metricool announced official connectors for both Claude and ChatGPT and opened them to all users, including the Free plan. The Terms simply have not caught up.

The direction is clear. What began as an Advanced-plan feature that also required Claude Pro is now a standard door into every account. That makes the MCP the cheapest way to put an AI assistant on top of Metricool data, because the API itself remains locked to Advanced and Custom. For a solo marketer on Free, the MCP is the only programmatic-feeling access available at all.

Connecting Claude, step by step

From the Connectors Directory

Metricool is listed in Claude's Connectors Directory, and this is the simplest route because it needs no Metricool API key. Connect once from Claude Desktop or Claude.ai and the connection applies across both, plus the Claude apps for iOS and Android. Metricool says setup takes about five minutes and needs no code, and it suggests the Claude Desktop app on Mac or Windows is where the MCP features are most powerful. Its own walkthrough is short:

1. In Claude, open the connector settings and find Metricool in the directory.
2. Click Connect.
3. Sign in to your Metricool account in the browser tab that opens.
4. Grant access.

As a custom connector, and in Claude Code

If you prefer to add it by address, the URL flow is the same in spirit: add a new MCP connection, paste ai.metricool.com/mcp, sign in to Metricool and grant access. Authorization is OAuth, and Metricool warns that granting it gives access similar to signing in directly, so only authorize tools you trust. On Claude Free you can use only one custom connector at a time, which matters if you already rely on another one. Claude Code is a different story: it connects from the command line and needs a Claude Pro or Max subscription, with Claude Code installed first. The help center also notes Claude Desktop can connect through mcp-remote.

Connecting ChatGPT

Metricool appears in ChatGPT as a verified app, so the basic path mirrors Claude: find the app, sign in to Metricool and grant access. Your ChatGPT account needs access to Apps, and which ChatGPT plans include Apps is set by OpenAI, not Metricool. Metricool publishes no eligibility list of its own and documents no separate iOS setup. The connection uses the same permissions as your Metricool account, and Metricool says your data is not stored in the AI model.

There is one wrinkle worth knowing. The native Metricool app inside ChatGPT runs an earlier version of the MCP integration, so it does not return direct clickable links back to the scheduled post in Metricool. The help center says ChatGPT users with a paid Metricool account can connect manually to ai.metricool.com/mcp to get the newer version, while the MCP itself still works on every plan, and other Metricool pages list ChatGPT in Developer Mode as a supported client. Through the app or the manual connection, the tools cover brands, analytics including ad accounts, and the planner: view, create, edit and, on Advanced or higher, send for review.

ChatGPT has one practical edge over Claude today. You can attach an image or video file directly to the chat and have it used in a post. In Claude that does not work yet, so media has to come from a link. An older Metricool guide to Claude does mention uploading a video file to the chat, but the newer help article on scheduling says attachments in Claude are not supported yet. Plan on links until that changes.

Connecting n8n and other automation tools

Most clients connect by URL and OAuth: ChatGPT, Make, Zapier, Smithery, Mistral Vibe, Cursor, TypingMind, Claude Desktop and Claude.ai all take the address and a browser sign-in. n8n is the exception. It authenticates with a header token, X-Mc-Auth, carrying your Metricool API key, rather than the URL-only OAuth flow. The API key lives in Account Settings under API, and the API is available only on Advanced and Custom; Metricool's limits page also says a token-based MCP connection that uses the API requires Advanced. By our reading, an n8n setup therefore needs an Advanced or Custom account, even though the MCP itself is open to every plan.

Other clients have their own conditions. Cursor's documented install path needs the Cursor Desktop app. TypingMind offers an OAuth path, and if you only see a token-based option there, that path uses the API and requires Advanced. Some setups pass a METRICOOL_USER_TOKEN and METRICOOL_USER_ID from account settings instead of signing in through a browser tab. Gemini CLI, like Claude Code, connects from the command line.

What the AI can do once it is connected

Reading brands, analytics and competitors

Metricool says the server exposes more than 20 tools. They fall into three areas: brands, analytics (including the analytics of connected ad accounts) and the Planner. On the read side, the help center lists tools to get available metrics, get analytics data including competitors, get the best time to post and get scheduled posts. Blog guides name get_brands, get_analytics, get_best_time_to_post, get_scheduled_posts, and competitor tools such as get_network_competitors. Competitor analysis reads only the competitors you already set up in Metricool. The supported networks are Instagram, TikTok, LinkedIn, Facebook, X, YouTube, Pinterest, Threads, Bluesky and Twitch, plus Meta Ads, Google Ads and TikTok Ads analytics.

Writing to the planner

The write side creates and updates scheduled posts. Tool names differ between Metricool pages: one help article calls them create_scheduled_post and update_scheduled_post, while other guides use post_schedule_post and update_schedule_post. Whatever the name, a post can be published immediately, scheduled for a date and time, or set to the best time from your account data. Posts created this way show up in the planner, where you can review and edit them, and the AI returns a direct link to the post inside Metricool. Repeating a post works through the MCP on any plan, with each date becoming its own scheduled post, while recurrence inside the planner itself needs Advanced or Custom.

What it cannot reach

The list of gaps is just as important. Metricool's help pages say the MCP does not cover the inbox, post comments, SmartLinks, Google Business Profile reviews, autolists, Canva or Drive files, or Ads campaign management ^[3]. For ad platforms it gives analytics only. You can ask about campaign results; you cannot launch or change a campaign. That leaves community management, link-in-bio pages and review replies where they were: in the Metricool app.

Two pages muddy this. The general setup article lists permissions that include the inbox, SmartLinks, reviews, Ads campaigns, autolists and Canva or Drive folders, and Metricool's integrations page says a connected AI can work with inbox messages, SmartLinks and campaigns. By our reading, those describe the access you grant at sign-in, not tools you can use: the ChatGPT, Claude, basics and FAQ pages all say those areas are not available through the MCP at the moment. Comments specifically are reachable only through the API, on Advanced and Custom, for the 50 most recent posts per network, and not for X, Threads, Pinterest or Bluesky.

The scheduling rules that trip people up

Scheduling from a chat feels effortless, which is why the fine print matters. Posts created through the MCP land in the planner like any other, so you can still open them there, check the media and fix the text before they go out, or ask the AI to change them later through the update tool. These are the rules Metricool documents for MCP scheduling ^[4]:

• Media must be a direct link. A public URL, a file in Google Drive linked to your Metricool account (Drive is a paid-plan integration) or Dropbox works; a page link such as a YouTube video or an Instagram post fails.
• Best-time suggestions cover six networks only: Instagram, TikTok, X, Facebook, LinkedIn and YouTube. Threads, Bluesky, Pinterest and Google Business Profile get none.
• Each network becomes its own post. That lets you schedule different times per network, which the planner cannot do in a single post.
• Network rules still apply. Instagram needs an image or video, YouTube needs a video and a title, and TikTok always needs a title plus an image or video. The per-network point deserves a second look on Free. A post scheduled to three networks becomes three posts, and each counts against the 20-post monthly limit once it publishes. By our arithmetic, a weekly announcement sent to three networks uses 12 of the 20 posts in a four-week month. That ceiling arrives fast. Starter removes the monthly cap, with unlimited publishing under the Fair Use Policy, so a busy calendar driven from a chat window is really a paid-plan habit.

When it will not connect

Metricool's troubleshooting page walks through the usual failures, and most come down to using the wrong method for the client. First check the address: the production endpoint is ai.metricool.com/mcp. Then match the method to the client. URL and OAuth clients want you to paste the URL, sign in to Metricool and grant access. n8n wants the X-Mc-Auth header with an API key. Claude Code and Gemini CLI connect from the command line.

Next, check the client's own requirements. Claude Code needs a Pro or Max subscription. Cursor needs the Desktop app. Claude Free allows only one custom connector at a time, so a second one will not attach. If the connection works but a request fails, your Metricool plan may be the limit, for example a Free account asking for analytics older than 30 days. Finally, to reset a broken connection, revoke access in the AI client (Metricool's help center says this is where it is removed; some Metricool pages also mention account settings), then reconnect. Support is by in-app chat or email on every plan.

MCP, API or Metricool Studio: which one to use

Metricool offers three ways to put data to work, and they suit different people. The MCP is for natural-language AI clients: you ask, it fetches or schedules. It runs on every plan and needs no code, although Metricool positions it for developers and power users because it needs some configuration. The API is for programmatic export and automation. It is Advanced and Custom only, uses an access token from Account Settings, and is the only route to inbox comments. Metricool Studio is the consumer option: a reporting engine inside Metricool that builds views from a written prompt. It is described as powered by Claude and shares technology with the MCP.

Studio has its own limits. It is not on Free, it works only on data the app already collects, and without the Advanced Analytics add-on you get two views of ten versions each. It is also free on paid plans only for a limited time and is announced as a future paid add-on. Pick the MCP to work from your AI client, Studio to stay inside the app, and the API to build. Our Metricool API guide covers the token, call limits and Zapier side in detail.

The bottom line

Metricool's MCP is one of the more generous AI connections in social media tools: official, open source and available on every plan, including Free. It reads brands, analytics and competitors and writes to the planner, which covers most of what a marketer wants to ask a chatbot. It does not touch the inbox, comments, SmartLinks or ad campaigns, and your plan's limits ride along, so a Free account still stops at 20 posts and 30 days of data. Connect Claude or ChatGPT through the directory, use direct media links, and step up to Advanced only if you need approvals or the API. If you are still deciding on a plan, start with whether Metricool's Free plan is enough.