Macha
Features Done-for-you setup Macha on your help desk Continuous Agent Evaluation Chrome Extension Website Chatbot Build with Sidekick Build with Claude Code & Codex Use Cases Integrations Pricing Documentation API Reference Building Effective Agents
Log in Book a discovery call
Done-for-you setup ZendeskFreshdeskGorgiasFrontIntercom

AI on your help desk,
implemented by our team

We analyze your help desk, pick the tickets to automate, build the agent and go live in safe mode. You just review the results.

Book a discovery call
Analysed, built and deployed by us
Safe mode first
Live within a week

Trusted by teams globally

5 and 5
AG Analytics
Babaa
Corkcicle
Crumb
Didriksons
France & Son
Luigi's Box
Lumin
myMuesli
Orderchamp
Sushi Bikes
Tuning Teile

How it works

You talk to us.
We do the implementation.

Setup is included in your plan, and your first ticket category is live within a week of the call.

  1. We take the brief

    We learn how your support runs, the rules your team follows and which tickets you want off their plate.

    Your part A 30-minute call

  2. We analyse your help desk

    We come back with the ticket categories worth automating, and the ones that should stay with a person.

    Your part Connect your help desk

  3. We build the agent

    We build its knowledge base from your help centre, website and past replies, connect the tools it needs, and test it on your past tickets.

    Your part Tell us your business rules

  4. We deploy it in safe mode

    It writes internal notes inside your help desk. Your team decides what goes out to customers.

    Your part Review the drafts and give feedback

  5. We tune it and move to the next category

    We fix edge cases, tighten the rules and switch on automatic replies once you are comfortable.

    Your part Say when
How the done-for-you setup works

What the analysis looks like

We tell you what to automate
and what to leave with humans.

Before we build anything, you get a breakdown of your ticket categories: how much of your volume each one is, which we would automate first, and which should stay with your team.

  • Categories ranked by volume
  • A clear call on each one: automate, or keep with a person
  • Where we would start, and why

Help desk analysis

Example: an online store, last 90 days

Where is my order Automate first
24% of tickets
Returns and exchanges Automate
13% of tickets
Address and order changes Automate
9% of tickets
Damaged or faulty items Keep with your team
6% of tickets
Complaints and chargebacks Keep with your team
4% of tickets

Damaged items need photos checked and a judgement call on each one, so we would leave them with a person.

How support teams use Macha.

Support leads on what changed once the repeat tickets stopped arriving.

Nicole Garelick

The implementation didn’t feel like “work”.

Macha runs the front line of our candidate and reference support in Zendesk, a few thousand tickets a month. It triages everything that arrives, answers what it can in our own approved wording, and passes the rest to my team.

Nicole Garelick

Director of Support · Crosschq

Crosschq
Saad Hashim

No guessing, and it escalates the moment it isn’t certain.

Macha runs our Rewards claims support in Freshdesk, a few thousand a month. When a customer says their cashback is missing, the agent reads the whole thread, looks the purchase up in our own system through a custom integration they built against our API, and answers on what it actually finds.

Saad Hashim

COO · Sprive

Sprive
Yvonne Panarin

They tuned each agent with us until it held up.

Macha runs the front line of our Zendesk across both our brands, around six thousand tickets a month in German, English, French and Dutch.

Yvonne Panarin

Customer Support Manager · myposter

myposter

We keep watching
after it goes live.

Every ticket the agent handles is scored against the rules you gave us. Did it verify the account? Did it escalate when it should have? We group the misses by rule and fix them with you.

Learn about evaluation

Tickets following your rules

86%

↑ 4% vs last week
Followed 312
Partly 48
Missed 11

Last 30 days

@refundAgent
371 tickets checked
Jun 18Jun 25Jul 2Jul 9Today

Needs attention · 22 tickets

Before approving a refund, ask the customer for the reason for return.

We caught it on July 2 and had it back the next day.

Built For Trust.
Designed For Privacy.

Your data is safe with Macha AI. We meet the highest security standards so you can focus on support without compromise.

Nothing goes to your customers until your team approves it.

GDPR Compliant

Fully aligned with global data protection and security standards.

No Training On Your Data

Your content is never used to train AI models, and is kept only for the periods set out in our Privacy Policy.

Frequently Asked Questions

Can't find what you're looking for? Talk to us

One call to walk us through your support and your rules, connecting your help desk, and feedback on the drafts while the agent runs in safe mode. We do the analysis, the build, the testing and the deployment.
Yes, end to end. We analyse your help desk and support tickets to learn how your team answers, build the agent, connect the tools it needs and put it live in safe mode. Nothing reaches a customer until your team approves it.
Yes. The analysis, the build and the deployment are all part of your plan. There is no separate implementation fee.
Days, not months. Most teams have their first category running within a week of the kick-off call. The slowest part is usually your review of the wording, not our build.
It stops and hands the ticket to your team with everything it has already gathered attached: what it looked up, what it found, and why it paused. It never guesses at a customer, and anything you tell it to always escalate goes straight to a person.
Zendesk, Freshdesk, Gorgias and Front. The agent works inside the help desk you already use: same queue, same views, same macros. There is nothing to migrate and nothing new for your team to learn. Tickets it handles get closed in your help desk like any other, with a full trail of what it did.
Any app your team already uses: Shopify, Stripe, Slack, Notion, Google Workspace and more, plus your own internal systems through a custom integration. It uses them to do the work, not just to look things up: pulling a live delivery date, issuing a refund, updating an address, checking a subscription. If a ticket takes three tools to resolve, it uses three.
Yes. We build a custom integration against your API so the agent can read from and write to your own systems, and that build is part of the setup rather than something you scope separately. Customers run their order and account look-ups this way today, so the agent answers on what it finds in your system instead of guessing from the ticket.
Yes. Each agent handles one category and has its own instructions, tools and approved wording, so they never bleed into one another. Most teams start with a single category, watch it for a couple of weeks, then add the next one once they trust it.
Nothing to start: $50 in credits, no time limit and no credit card. That is enough to run a real category against real tickets, so you can judge it on your own queue rather than on a demo.

Put AI on your first ticket category. We'll build it.

Tell us which tickets your team answers over and over. We analyse them, build the agent, put it live in safe mode and keep it tuned as your tickets change.

Book a discovery call
Analysed, built and deployed by us
Safe mode first
Live within a week