Learn

Agents: let ChatGPT, Claude and others work on your listsConnect ChatGPT and Codex to RomaConnect Claude and Claude Code to RomaConnect Cursor, VS Code and Gemini CLI to RomaConnect Poke, Town, Grok and other agents to RomaAPI keys and signing in from an agentWhat an agent can and cannot do in your accountTriggers: wake your agent when something changes

Connect Claude and Claude Code to Roma

Add Roma once from Claude's directory, or paste one address by hand.

Claude connects to Roma from Claude's own directory: add Roma there, sign in and approve once, and Roma is in Claude on the web, in Claude Desktop, in Claude on your phone, in Cowork and in Claude Code, everywhere that account is signed in. Claude Code can also add it from the terminal with one command. The page for both is /agents/claude, and Agents in Roma explains what a connected agent is.

From Claude's directory

Roma is listed in Claude's directory at claude.ai/directory/roma. Open the listing, add Roma, sign in to Roma with Google or Apple, and approve. That is the whole setup, and the one to use: the listing covers Claude on the web and your phone, Claude Desktop and Claude Code at once.

To check it, ask Claude "What's on my plate today in Roma?" The answer should use your own task titles.

As a custom connector

If you would rather add Roma by hand, or under another account, connect it once on claude.ai and it is available everywhere you are signed in:

  1. Open Customize, then Connectors, then Add, then Add custom connector.

  2. Name it Roma, paste https://api.roma.app/mcp, leave sign-in on, and add it.

  3. Choose Connect, sign in to Roma, and approve.

A Free account can add one custom connector. The paid plans add as many as you like. On a Team or Enterprise plan an owner adds Roma under Organization settings, then Connectors, then Add, then Custom, then Web. Each member then finds it under Customize, then Connectors, and chooses Connect.

Claude Code, from the terminal

The directory listing and a connector added on claude.ai both show up in Claude Code. To add it from the terminal instead:

claude mcp add --transport http roma https://api.roma.app/mcp

Then run /mcp inside Claude Code and choose Roma to sign in. When Roma adds a tool, Claude Code picks it up in the running session. On a machine with no browser, Claude Code takes an API key as a header instead of the sign-in: API keys and signing in from an agent says how.

Using Roma in Claude

In a conversation, Roma is on under the plus menu's Connectors. The first time a tool runs, Claude asks and offers Allow always for that tool, so the writes can be allowed once and never asked about again. On a Team or Enterprise plan an owner can also set each tool to Always allow, Needs approval or Blocked, under Organization settings, then Connectors, then Roma, then Tool permissions.

Then talk to Claude the way you already do. Roma is there whenever a task comes up: "What's on my plate today?", "Add 'call the dentist' and 'send the invoice', both for Friday.", "Save the plan we just made as a note." What an agent can and cannot do in your account lists everything it reaches.

Make Roma the list Claude reaches for

Send Claude the setup message from /agents/claude once. Claude makes Roma its to-do list, looks through your recent conversations for what is still open, shows you the list, adds it to Roma after you confirm, and keeps Roma current as you work.

For a standing line that holds across every conversation, two things help. Put the instructions line from /developers/connect into Claude's instructions, so each conversation starts from where you are. And set Roma's read-only tools to Always allow under Customize, then Connectors, so Claude can look at your lists without asking each time.

Claude on a schedule, or woken by a change

Claude can keep Roma current without you in the chat. In Cowork, type /schedule in any Cowork task, pick hourly, make sure the Roma connector is on, and paste the standing order from the Recipes page at /developers/recipes. Every run it looks through your meeting notes, calendar and unread mail for commitments and dates, adds what is new to your Tasks list, and works the Roma tasks you hand it, continuing in each task's body and posting in its chat only when something needs you. A Claude Routine does the same from the cloud, with Roma as its connector.

A trigger goes one step further: Roma starts a Claude Routine the moment something on a list changes, with no schedule at all. Triggers: wake your agent when something changes explains how.

Questions people ask

Do I have to add Roma again in Claude Code if I added it on claude.ai?

No. A connector added on claude.ai, from the directory or by hand, shows up in Claude Code too, under the same account. The terminal command is for adding it under another account, or on a machine where you have not signed in to claude.ai.

Will Claude ask me every time before it changes something?

Only the first time per tool, unless you want it to. When a tool runs for the first time, Claude asks and offers Allow always for that tool. Set the read-only tools to Always allow as well, and looking at your lists never asks either. Every change still shows in your Roma Event Log as your own action, made through Claude.

Can I use an API key with Claude instead of signing in?

Claude on the web, the Mac and the phone does not take a pasted key and signs in with your Roma account in the browser. Claude Code does take a key, as a header, for a machine with no browser. API keys and signing in from an agent says how to create one.

What if the sign-in window does not open?

A pop-up blocker, an ad blocker or Safari is stopping it. Tell Claude it did not open and ask for the sign-in link as plain text, then open it in a tab of your own. The longer list of what can go wrong is on /developers/connect.