I’ve been answering support tickets for 20 years.
Well, not "me" but someone in my company, often my wife, usually a Filipino assistant.
The first thing everyone thinks when you say "automating customer support" is great, this companies dead to me their support is going to suck now.
It's a fair response.
We've all been on those horrible phone calls on hold, or those phone tree systems where you have to press a number to get to the right person.
Here's what people don't understand though...
80% of customer support is helping people recover lost passwords, finding them the link to a file, and there's 12-20 common questions that get asked on repeat. Email inbox support is much different than phone support. Plus, we can always stil escalate issues to a human, and should.
Last month I built an AI agent that handles all of it. The junk gets cleaned out before I wake up. Real customer emails get drafted responses that land in my Telegram. I review them, say “send it,” and I’m done. 5 to 10 minutes a day instead of an hour.
Well... again, not "me" someone in my company, often my wife lol
We used to pay as much as $2,000/mo for customer support reps and now we pay about $60/mo in API fees for our customer support agent.
The best part... people are loving Jeffri! (our support agent)
He's fast, he's curteous, he never catches an attitude, and he's knowlegable (trained on years of email history).
Here's how I did it...
The Perfect Customer Support Agent
There are 3 tools that when combined create the perfect combination of technology for a customer support agent.
Hermes: The key part of Hermes is that you can install it on your local computer or your own private cloud server. This is critical when handling customer data. Their are laws about storing and handling customer data, but since Hermes can install locally you can mitigate that risk.
Learn more about Hermes at: https://hermes-agent.nousresearch.com
Gbrain: This is a free tool made by the founder of Ycombinator, a famous startup incubator out in Silicon Valley. What makes it so great for customer support is it's purpose built for relationship management. It will create a memory page for every single customer so your Hermes agent can "remember" every interaction it's ever had with that customer, their issues, purchases, cancelations, etc.
HelpScout MCP: We've been using HelpScout for years as our customer support ticket inbox software. With the MCP it allows our AI agents to login on our behalf, send/receive emails, close tickets, open tickets, and more. This is the little bridge that makes the whole thing possible.
The Easy Way to Build This
Here’s where most non-technical founders bail out. “I need a server” sounds like “I need a computer science degree.” It’s not.
DigitalOcean (a hosting company) has a 1-click install for Hermes. That means you can spin up a server with Hermes pre-installed in about 90 seconds. You don’t open terminal. You don’t install packages. You click a button and it’s running.
But the real discovery was the CLI.
CLI stands for Command Line Interface. All you need to know is CLI is how computers talk to each other. Computers don't login, scroll, and click buttons like we do.
No, they just enter commands and stuff gets done.
That's why it's called "Command... Line... Interface!"
Here’s why you’re going to love CLIs.
Because Claude, Manus, and ChatGPT already knows how to use them. When a service has a CLI, you just tell Claude what you need, and Claude can just login and do it for you.
DigitalOcean has a CLI.
It's called "DOCTL" 🤷♂️ (don't ask me). Every piece of this build was “tell Claude what I need, do what Claude says next.” That’s the whole workflow.
I didn’t even know how to install Hermes. I told Claude “I need Hermes Agent installed on a DigitalOcean droplet” and Claude walked me through every step. SSH into the server. Check the version. Configure the API keys. Set up the Telegram bot. All of it.
If you can take a screenshot and type “what’s next,” you can build this.
🔑 MAJOR KEY... Before you do this make sure you enable "Control My Mac" and "Filesystem" in Claude customization settings. These two features allow Claude to use your computers terminal for you. Basically, you're giving Claude root access.
There Was Just One BIG Problem
A few days after setting up Hermes, Gbrain, and access to my Helpscout account I opened Claude in Cowork mode and said (roughly):
“My Hermes agent can receive HelpScout tickets but doesn't seem to be able to take any action on them. I need a way for the agent to read conversations, send replies, close tickets, mark spam, and manage the inbox. Build me whatever is needed to make that work.”
Claude knew exactly what to do. It told me I needed an MCP server. A small program that wraps the HelpScout API and exposes it as tools that Hermes can call.
Like giving the agent hands to go with the ears it already had.
I have no clue how to build an MCP server, I just know that MCP is how tools connect to AI.
That's ok, because Claude did.
In about 20 minutes, Claude had written the entire MCP server from scratch. One Node.js file. 10 tools. Full inbox control. Authentication that handles itself (tokens refresh automatically so nothing breaks at 3am). Clean error handling. Production-ready code.
The 10 tools Claude built:
- list_conversations - Pull active tickets from the inbox
- get_conversation - Read the full email thread
- reply_to_conversation - Send a reply to the customer
- add_note - Add a private internal note
- close_conversation - Close a resolved or irrelevant ticket
- mark_as_spam - Flag and close spam
- delete_conversation - Permanently remove a conversation
- update_conversation - Change subject, status, or assignment
- tag_conversation - Add tags for categorization
- search_conversations - Search by subject, email, tag, or criteria
I'm going to give you the full MCP files Claude made for me in the Build Notes+ premium section.
But here's the next problem...
You have to host the MCP server, ON A SERVER. Good news, already got one of those with DigitalOcean. Claude already knows how to login and install it for me.
Claude installed everything. Claude wrote custom code to make my setup work. Claude then installed the code for me.
I don’t know how APIs work under the hood.
I couldn’t have written that MCP server if you put a gun to my head.
It didn’t matter.
Any frontier model (Claude, ChatGPT, Gemini) has already read the entire internet. Every API documentation page. Every Stack Overflow thread. Every YouTube transcript for every tutorial ever recorded.
They can search the docs for any tool.
They can read the source code for any open source project. They can figure out the authentication flow for any service.
You don’t need to know how. You need to know WHAT.
What do I need? Full inbox control. What tools are available? HelpScout API. What agent framework am I using? Hermes with MCP. What behavior do I want? Emails drafted and waiting for me.
AI figures out the rest.
Before all of this customer support took us 2hrs a day and cost me $2,000/mo to have multiple support reps manning it from different time zones.
Now, it costs me $60/mo and takes less than 10mins a day. With customers that are happy about the change.
Now, let me give you the prompts to type in and the MCP server files that Claude built for me, so you can rebuild this tomorrow...
Continue with Build Notes+
This is the free preview. The complete step-by-step guide and supporting files are available to Build Notes+ members.
