Agents: let ChatGPT, Claude and others work on your lists

An agent in Roma is an AI you already talk to, ChatGPT, Claude, Cursor or another, connected to your account so it can read and write your lists. You connect it once, in a browser, with the same sign-in you use for Roma. From then on, whatever you ask that agent to keep lands in Roma, where the operator, your other agents and you on every device see the same lists.

One list behind every AI you talk to

Mara plans a trip with Claude on her phone and says "remind me to renew the passport on Friday". The task appears on her Tasks list in Roma, due Friday, before she has put the phone down. That evening she asks Cursor, in the middle of some code, what is on her plate, and it answers from the same list. Nothing lives only in one chat.

That is what connecting an agent means. The agent sees what is due, adds what you mention, and knows what is next, from your real list rather than from the conversation alone. Roma stays the one place your tasks, notes and lists live.

Which agents connect

Roma has a page for each agent at /agents, with the steps and a message to copy:

• Claude and Claude Code, from Claude's own directory in one click, or by hand. Connect Claude and Claude Code to Roma.

• ChatGPT and Codex, on chatgpt.com and in the terminal. Connect ChatGPT and Codex to Roma.

• Cursor, VS Code and Gemini CLI, with a one-click link or one command. Connect Cursor, VS Code and Gemini CLI to Roma.

• Poke, Town, Grok, Muse, Instinct, OpenClaw, Hermes Agent and Devin Desktop. Connect Poke, Town, Grok and other agents to Roma.

An agent that is not listed connects too. Every agent reaches Roma through one address, https://api.roma.app/mcp, and any agent that speaks MCP, the open standard these connections use, can be pointed at it. An agent with no settings screen at all, one that lives in your messages, connects from the message alone.

How a connection works

The first time an agent connects, a browser opens Roma's sign-in. You sign in with Google or Apple, then approve the agent on a screen that names the app and says where it returns you afterwards. Nothing is pasted and no key is copied. The agent keeps its own sign-in from then on and renews it by itself.

An agent that cannot open a browser, a scheduled job or a self-hosted agent, uses an API key instead, created under Settings, then Connections, and pasted where the agent asks for one.

The message that does the setup

Every agent page carries one short message to copy and send to the agent. It says, in your own words, that you use Roma for your tasks, where the ways to connect are explained, and asks the agent to try the way most likely to work for it and to tell you what it is doing. Most agents set themselves up from it. An agent with a settings screen takes the steps on its page instead.

Once connected, a second message on the same page makes Roma the agent's to-do list. The agent looks through your recent conversations for anything still open, shows you the list, adds it to Roma after you confirm, and from then on adds what you mention and answers "what should I work on next" from Roma. Then talk to it: "What's on my plate today?", or "Save the plan we just made as a note."

What an agent can do, and what it asks first

A connected agent can see what is due, add tasks with their due dates, save a note, fill a list with many items at once, search everything you keep, read the chat on an item and leave a note there under its own name, read your automations and start one, and bring something back from the trash. It acts only as you, on your own workspace, plus what others shared with you.

Every tool an agent gets says whether it reads or changes something. An agent runs the reading tools freely, or asks once and remembers, and asks before anything changes: Claude asks the first time a tool runs and offers to allow it always, ChatGPT confirms and can remember your choice for the rest of the conversation. Starting an automation always asks, because a run can act through your connected apps. What an agent can and cannot do in your account has the whole picture, including the 30 days every deleted item spends in the trash.

Everything an agent changes shows in your Event Log as your own action, marked as coming through that agent, and every edit to a body is kept in version history. The Event Log is where to look when you wonder what an agent did.

Agents, connections and automations

Three different things meet on the Connections page of Settings, so it helps to tell them apart. A connection is an app Roma reaches into for you, Gmail or Google Calendar, say. An agent is the other direction: an AI reaching into Roma. An automation is a standing order Roma runs on its own. An agent can read your automations and start one, but creating or changing one stays in Roma, where an approval card shows you the schedule and the steps.

Roma can also wake an agent the moment something on a list changes, instead of the agent checking on a schedule. That is a trigger, set up on the Automations page: Triggers: wake your agent when something changes.

Where the detail lives

The agent pages at /agents are all most people need. The developer docs at /developers have the rest: every tool with what it takes and returns, the sign-in flow, the standing orders to paste into a scheduled agent, and the triggers.

Questions people ask

Does a connected agent see other people's things?

No. An agent acts as you and reaches your own lists and the items on them, plus the items and lists other people shared with you, with the rights you have there. It does not reach your connected apps or your email through Roma, and nothing of other people's beyond what was shared with you.

Do I need a paid plan to connect an agent?

On Roma's side you need a Roma account and nothing else. Some agents limit custom connections on their side, ChatGPT's Developer mode for one, and each agent's article says what applies.

How do I disconnect an agent?

Remove Roma in the agent's own settings. That deletes the sign-in the agent holds. Signing out of Roma does not disconnect an agent, and disconnecting an agent does not sign you out of Roma. To have the grant revoked on Roma's side as well, write to hello@roma.app.

Can I connect more than one agent?

Yes. Every agent connects on its own to the same account, through the same address, and each one sees the same lists. Connect the ones you use.