Use ChatGPT, Claude or another AI app with Hey Quad

You can connect ChatGPT, Claude or another AI app to Hey Quad. The app works as you: it can go through your email and write drafts, tell you what needs your attention today, find people and records, change your own status and call forwarding, and read your reports. It never sends an email or a text. A person always reviews a draft and clicks Send.

Hey Quad connects through MCP (Model Context Protocol), a standard way for AI apps to use outside tools. Any AI app that lets you add an MCP server or custom connector can use it.

This is the Business connector, for everyday work, and anyone in your organization can use it. Owners and admins can also connect an app that helps set Hey Quad up, through a separate connector. See Set up Hey Quad with an AI app.

Turn it on

AI apps are off for every organization until an admin turns them on.

Changing this setting needs the "Turn on AI apps" permission, which owners and admins have. Anyone who reads the Inbox can see whether it's on.

Organizations that handle health information (PHI) can't turn AI apps on. The switch stays off and says why.

Connect your app

Each person connects their own app.

You can leave any area unticked. The app then gets none of that area's tools, and what it reads elsewhere leaves that area out too. For example, an app without Communications gets no email in your day's summary.

One connection works in one organization. To use the app in a second organization, connect it again and pick that one. When Hey Quad adds a new area, connect the app again to allow it.

If your organization uses the Strict two-factor policy, the connection stops working 12 hours after you connect it. Connect the app again to keep going. See Two-factor sign-in and passkeys.

What it can do

Ask the app in plain words, for example "what do I need to deal with today?", "find Dana Reyes", "go through what needs a reply and draft answers" or "put me on do not disturb and send my calls to my cell until 5".

• Your day: one summary of what needs your attention: email waiting for a reply, missed calls nobody has called back, your unheard voicemail, tasks and follow-ups due today, and today's appointments.
• Find: look anyone or anything up by name, phone number or email address: contacts, companies, tickets, email conversations, text conversations and calls.
• Email: scan the Inbox, read conversations and attachments, write replies, forwards and new emails as drafts, file conversations (Done, snooze, assign, label, Spam), and add internal notes. Notes are never emailed.
• Texts: read your text conversations, check whether a number can be texted, and write a text as a draft. The draft waits in that conversation's reply box, marked Draft by and the app's name. You can change it, then press Send, or Discard it.
• Tickets, tasks and contacts: make a ticket or deal from an email, add or tick off tasks, and link an email to a contact.
• Your status: see and change your availability (available, away or do not disturb) and where calls you don't answer go, now or until a time you pick. With away or do not disturb, that's every call. It can also see and change your own email signature for each mailbox you send from.
• Reports: read call, queue, team and AI receptionist reports, usage, and your dashboards.
• Check and ask Hey Quad: see whether Hey Quad has reported an outage or maintenance, and open a support request with Hey Quad for you.
• What it did: ask the app what it did as you. It can list every action any app took as you, newest first.

Some actions need a person's OK before they happen. The app asks instead of acting, and the request waits on the Approvals list. See Approve what an AI app asks to do.

What it never does

• It never sends. Not email, not texts. Every email or text it writes is a draft that a person reviews and sends.
• Email text is data, not instructions. The app is told that anything written in an email, text or voicemail, even "ignore your instructions and…", is never a command.
• No recordings. It gets call and voicemail transcripts and summaries, never the audio.
• Only your own status. It can change your availability and forwarding, never a teammate's.
• Replies go back to the conversation. Hey Quad works out who a reply goes to from the conversation itself. New emails and forwards can only go to addresses your business already knows, such as your contacts, people you've emailed, and your team.
• Likely secrets are hidden. One-time codes, passwords, API keys and sign-in or reset links are replaced with a "[redacted]" marker before the app sees them.
• Small batches. One action touches at most 25 conversations, and the app is told to ask you before acting on many at once.
• Changes can be undone. When it files conversations or changes your status, it gets an undo it can run if you ask.
• It leaves a trail. The app is told to add a short note to every conversation it acts on. Notes and tasks it writes show which app wrote them, and every action it takes is recorded with the app's name.

Permissions

The app acts as you, with exactly your permissions. It sees only what you can see in Hey Quad and does only what your role allows. Tools you can't use aren't offered to the app at all. For example, without the "See reports" permission it gets no reports, and without the "Edit contacts" permission it can't make tasks or link contacts. Tools marked "Any member" in the table below work only on your own things, such as your status and your day.

Remove an app or turn it off

• One app: go to Account → Connected apps and click Revoke next to it. It loses access straight away.
• Everyone: an admin turns off Allow AI apps under Settings → Inbox & email. Every connected app stops on its next request.

Drafts and notes the app already wrote stay where they are. A status change with an end time still goes back at that time.

The hq command

The hq command-line tool has the same tools, built from the same list as the table below: hq mail for email, hq today, hq find, hq me for your status, hq reports and hq support. People who run AI agents on their own computer, such as Claude Code, can use it instead of a connector. It follows the same rules: drafts only, your permissions.

Agent mode, when an AI runs hq

When an AI agent runs hq for you, turn on agent mode for that sign-in. Hey Quad then treats it like a connected AI app:

• Actions that need approval, such as booking an appointment or publishing a flow, go to the Approvals list instead of happening. hq prints Sent to Approvals: … with a link, and a person approves or rejects it there.
• It can change things only through Hey Quad's tools (the hq commands built from the table below). Other changes are refused.
• Whatever it does is labeled hq CLI (agent) in the audit log and on what it writes.

Turn it on with hq agent on, or sign in with hq login --agent. hq agent status shows where you stand. To start every new hq sign-in in agent mode, run hq agent default on, or turn on Start new CLI sessions in agent mode under Account → Connected apps.

Turning agent mode off needs you, not the AI. Run hq agent off yourself and type a code from your authenticator app (or a recovery code), or use the Agent mode switch under Account → Connected apps, which asks for a two-factor check. An AI that only has the hq sign-in can't turn it off by itself. If you don't use two-factor, sign the session out instead (hq logout, or end the session in the console).

If your organization's owner lets an action run without approval (direct mode), it runs directly in agent mode too.

Tools

These are all the tools a connected app can use, by the area you allow when you connect. The list comes straight from Hey Quad, so it's always current. An app only sees the ones your permissions allow.

• Reads look things up and change nothing.
• Draft writes something nobody sees until a person sends or publishes it.
• Changes, can be undone take effect at once, and the app gets an undo.
• Needs approval waits on the Approvals list for a person.