Triage and Label a Cold Email Inbox with AI

How do you triage and label a cold email inbox with AI?

HotHawk already labels every reply with its own AI, into buckets like Interested, Not interested and Out of office. Want your own logic? Turn the built-in labelling off and point Claude or ChatGPT at the inbox over MCP or the API to sort it to your rules. Labels let you filter the inbox and can trigger a follow up.

Start your 7 day free trial Connect the MCP server

No credit card required.

Before you start

You need a HotHawk account and the HotHawk MCP server connected to Claude. It takes a couple of minutes and no code. Full setup is on the MCP server page.

Step by step

1

You ask Claude
What labels is HotHawk already putting on my replies?

Claude calls mcp__hothawk__labels_list

workspaceId current workspace

HotHawk already sorts every reply out of the box with its own AI: Interested, Not interested, Out of office and more. You can add your own labels on top of the defaults.
2

You ask Claude
Switch the labelling over to Claude and use my own rules.

Claude calls mcp__hothawk__workspaces_settings_update

aiLabelling off

The built-in labelling steps aside so your model can take over. Point Claude at the inbox over MCP, or your own agent over the API, when you want the sorting done to your exact definitions.
3

You ask Claude
Read this week's replies and label them by my rules.

Claude calls mcp__hothawk__threads_update

threadId each unlabelled thread label matched intent

Claude works down the Master Inbox and labels each reply by what it actually says. Now you can filter the inbox to just the Interested ones, or let a label trigger a follow up.

Your inbox can label itself

Send at any volume and the replies pile up fast: real interest, soft maybes, hard nos and a wall of out of office bounces, all arriving in the same place. The good news is HotHawk does not leave you to sort that by hand. Its own AI reads every reply as it arrives and labels it, out of the box, into buckets like Interested, Not interested and Out of office. You can add your own labels on top.

So from day one your inbox triages itself. You open the Master Inbox and the replies are already sorted, before you connect any model of your own.

Bring your own AI when you want more

The built-in labelling is quick and it runs on HotHawk, so nothing leaves the platform. When you want more accuracy, or a sort that follows your exact rules, you can take it over. Turn the built-in labelling off in your workspace settings, then point your own model at the inbox: Claude over the MCP server, or your own agent over the REST API.

Now the labelling is yours. Tell Claude what each label means for your offer, and it reads each reply and sorts to that. A "sounds good, but not until Q4" gets Not interested for now with a note to nudge later, rather than a blunt Positive. You get the judgement of a frontier model on your own definitions.

What the labels are for

Labels do two jobs. First, they filter: pull up every Interested reply in one view, or every Out of office, so you work a queue instead of a pile. Second, a label can trigger a Sub Sequence, so tagging a thread can set off the follow up on its own. Deciding which rep a reply is assigned to is a separate setting, so a label sorts the inbox, it does not quietly reroute your team's work.

Because the Master Inbox catches every reply across every mailbox, including the forwarded, CC'd and out of office ones, there is nothing hiding outside the sort. Whether HotHawk's AI is doing the labelling or yours is, the inbox stays current and the labels keep everything downstream moving.

The HotHawk tools behind this

Every action here is a real endpoint. Claude reaches it over the MCP server, and you can hit the same endpoints directly from the REST API if you would rather build it into your own stack.

mcp__hothawk__labels_list See the labels HotHawk already applies, and your own
mcp__hothawk__workspaces_settings_update Turn the built-in labelling off to use your own AI
mcp__hothawk__search_threads Pull the replies to work through, or filter by label
mcp__hothawk__threads_update Apply a label to a thread

MCP server setup Read the API docs

Frequently asked

Does HotHawk label my replies automatically?

Yes, out of the box. HotHawk sorts every reply with its own on-platform AI into default labels like Interested, Not interested and Out of office. You can add your own labels too. So your inbox triages itself from day one, before you connect anything.

Why point my own AI at it instead?

For more accuracy or your own logic. If you want the labelling done to your exact rules, turn off the built-in AI and label with your own model over MCP or the API. Claude or ChatGPT reads each reply and sorts it the way your team actually works.

What do the labels actually do?

Two things. They let you filter the inbox, so you can pull up every Interested reply or every Out of office in one view, and a label can trigger a Sub Sequence. Deciding which rep a reply goes to is a separate assignment setting, not the label.

Will it get labels wrong?

It sorts the clear ones and you can have it flag anything borderline back to you rather than guessing. You can also have it draft the labels for your approval and apply nothing until you say so. You set how much rope it gets.

Do I need to write code?

No. The whole flow runs in plain language over the MCP server. The label and thread actions are in the REST API too if you want to wire your own labelling into your stack.

Keep going

Find and tag hot leads Surface the strongest buying signals in your inbox Auto follow up after a reply Turn a label into a follow up that fires on its own AI email assistant Let HotHawk sort and answer your cold email replies AI cold email The whole picture of AI cold email on HotHawk

How to triage and label your cold email inbox