Product.

Everything a support team needs, nothing you can’t change

Hatch is a complete help desk out of the box. When your process outgrows the defaults, you change the desk instead of working around it.

The desk.

Eight parts, one data model

Every feature reads and writes the same tickets, customers and events, so reports, rules and AI all agree on what happened.

01 Inbox

One inbox for every channel

Email, chat, forms and Slack land in the same queue, threaded by customer, with the full history beside every reply.

02 Queues

Queues and SLAs

Route by product, plan, language or anything in the ticket. Timers show what breaches next before it does.

03 Automations

Rules without scripts

Tag, assign, escalate and close on a trigger. Every rule has a dry run so you see what it would have done.

04 AI replies

Drafts from your own docs

The reply engine reads your help center and past tickets, cites where an answer came from, and never sends without a person.

05 Help center

Articles that answer first

Publish a help center on your own domain. Search deflects the common questions before they become tickets.

06 Reporting

Numbers the team trusts

First response, resolution time, CSAT and backlog by queue, built on the same tables you can query yourself.

07 API and MCP

Typed API and MCP server

Every object is in the API. Claude, ChatGPT and Cursor can read tickets and draft replies through MCP.

08 Extensions

Extend it in TypeScript

Add ticket types, actions and integrations in your own repo, reviewed and deployed like the rest of your code.

Skip the clunky UX that custom usually means.

Give your agents a desk they'll love opening

Calm, familiar inbox

Tickets sit on a board by status. Drag one to move it, the way your team already works.

Live and multiplayer

Everyone sees the same queue update as it happens, so two agents never answer the same customer.

Every action a keystroke

Press ⌘K to reply with a macro, merge duplicates or jump to any queue without reaching for the mouse.

A Hatch ticket panel with fields, timeline and an AI draft, in halftone bars

AI that shows its work.

Agents check a reply instead of writing one

Hatch drafts answers from your runbooks and past resolutions. Each draft links the sources it used, so a wrong answer is easy to spot and fix at the source.

  • Runs on your model

    Bring Claude, GPT or a model you host. Keys and prompts stay in your workspace.

  • Learns from edits

    When an agent rewrites a draft, the change becomes a suggestion for the article it came from.

Read the agent guide

Built to run.

Fast at ten tickets, steady at a million

<90ms
Median ticket open time on Hatch Cloud
99.98%
Cloud uptime over the last 12 months
40+
Integrations maintained by the core team
1 hr
Typical time to forward email and go live

Get started.

Open up your support desk

Start on Hatch Cloud in minutes, or run it on your own servers for free.