MCP FAQ: Connecting Routinery to ChatGPT and Claude (2026)

Last updated: 2026-09-20

Quick answer: Routinery's MCP server is free for every Routinery account, so the requirement that varies is the plan on your assistant. Claude allows custom connectors on all plans including Free, capped at a single connector on Free. ChatGPT needs a paid plan with Developer mode on. Cursor and Gemini CLI work on all plans.

Two guides on this blog already walk the click paths, one for Claude and one for ChatGPT. The questions that keep landing afterward sit between them. What plan do I need, and can the assistant see anything before I've signed in? Why does it work in one chat and go blank in the next? Those answers are scattered across Routinery's MCP page, Anthropic's connector help center and OpenAI's connector docs, so they're collected here.

Plan requirements: what each side of the connection needs

Routinery charges nothing for MCP at any subscription level, which leaves your assistant's plan as the only thing that can block the connection.

One wrinkle on the ChatGPT side stays out of the table. OpenAI's documentation notes that Developer mode availability can depend on account and workspace policy, so a paid seat inside a company workspace can still come up short. Turning the mode on takes 3 steps under Settings, and adding the server afterward is a separate 6 step form. Anthropic's limit is a count. Free Claude accounts are limited to 1 custom connector, so Routinery has to be the one you keep, or the plan moves up.

Access and approval: what the connection can reach

Access begins at the sign in. Nothing's visible before that. Routinery's MCP page puts the mechanism in one line, that the server uses OAuth 2.0 for authentication and an assistant can read your routines only after you sign in and approve.

Writes are gated a second time. Routinery asks for explicit confirmation before it creates, duplicates or edits a routine, and you'll review the exact action before it saves. The assistant drafts the change in the chat and stops there.

Your routine data goes to the assistant you picked. Routinery's page names the companies it can be, OpenAI or Anthropic or Google, depending on the client you connected. Anthropic makes the same point from its own side of the wire, warning that custom connectors link Claude to services Anthropic hasn't verified and that connecting one grants Claude the ability to access and potentially modify data within those services based on your permissions. Routinery's connector isn't exempt from that warning.

Enabling it inside the chat: the step that reads like a broken connection

Adding a connector in settings and switching it on inside a conversation are separate actions, and skipping the second produces an assistant that behaves as though it's never heard of the app. Both clients document where the switch lives. In Claude you open the plus button at the lower left of the chat and pick Connectors. ChatGPT keeps the same control in the tools menu of a new conversation. Per conversation means what it says. A fresh chat starts from off.

The initial setup differs enough between the two clients that each one has its own walkthrough. On Anthropic's app, connect Routinery to Claude step by step covers the connector screen and the sign in. The OpenAI path is in build a routine with ChatGPT via Routinery MCP, which also covers Developer mode and the first prompts worth trying.

Removing the connector: where the Claude setting lives

Claude's help center documents the removal path directly. Open Customize, then Connectors, find the Connectors section, and click Remove or the three dots beside the connector you want gone. The prompts finish it. Putting it back is the same 5 steps you used the first time on a Pro or Max account, so nothing about removing it is permanent.

Frequently asked questions: Routinery MCP on ChatGPT and Claude

Which plans support Routinery MCP on ChatGPT and Claude?

Routinery's own side is free for every account regardless of subscription status. The answer depends on the assistant you connect. Claude allows custom connectors on Free, Pro, Max, Team and Enterprise plans, with Free accounts capped at a single custom connector. ChatGPT needs a paid plan, meaning Plus, Pro, Business or Enterprise, with Developer mode switched on, and OpenAI adds that the mode's availability can depend on account and workspace policy. Cursor and Gemini CLI work on all plans. You'll need a Routinery account either way, since the first connection attempt sends you through an OAuth sign in before any routine is visible to the assistant.

Can an assistant change my routines without asking?

No. Routinery requires explicit confirmation before creating, duplicating or editing a routine, and the page asks you to review the exact action before approving it. In a chat that shows up as the assistant writing out the change it intends to make, then waiting. Reading works the same way one level up, since nothing about your routines reaches the assistant until you've signed in through OAuth and approved the connection. That puts two approvals in the path. The connection needs one, and each proposed change needs another.

What data does Routinery share with ChatGPT or Claude?

Your routine data. Routinery's MCP page states that connecting through MCP shares your routine data with the AI platform you selected, naming OpenAI, Anthropic and Google as the examples depending on the client. Anthropic's help center describes the same arrangement from its end, warning that connecting Claude to an external service grants it the ability to access and potentially modify data within that service based on your permissions. Both statements point at the same trade. The assistant has to read your routines to be useful with them, and that read puts them on its servers.

How do I remove the Routinery connector from Claude?

Anthropic's help center describes going to Customize, then Connectors, locating the Connectors section, and clicking Remove or the three dots next to the connector before following the prompts. Adding it back later is the same five step form on Pro and Max accounts, so the removal isn't a one way door. ChatGPT works the same way from its settings menu, where a connected MCP server can be switched off, though OpenAI's published documentation covers connecting rather than removing, so that one comes from the screen rather than a page we can link. Turning a connector off for a single conversation does something else. That one happens from the plus button inside the chat, and it leaves the connector installed for every other conversation.

My assistant says it can't see my routines. What should I check?

Start with whether the connector is switched on in the conversation you're using, because adding it in settings doesn't enable it in a chat. Claude keeps that control behind the plus button at the lower left. On ChatGPT it lives in the tools menu of a new conversation, and the setting resets per chat in both apps. If the connector never appeared as an option to begin with, the plan requirement is the next thing to look at, since ChatGPT puts custom connectors behind Developer mode on a paid plan and Anthropic holds Free accounts to a single custom connector.

About the author. Written by the Routinery team. Every requirement on this page is checked against Routinery's own MCP page, Anthropic's custom connector help center and OpenAI's connector documentation, all linked in place so you can check them against your own screen before connecting anything.