Hermes Agent on AWS - Your Own Agent, Without Someone Else Metering Every Message

Most AI agent products charge you for the privilege of talking to your own data. Per seat, per message, or a markup baked into every token that passes through their servers. Hermes Agent doesn't work that way, because it isn't that kind of product - it's MIT-licensed open-source software built by Nous Research, and you can run it on hardware you control for less than the cost of a coffee a month.

Meetrix packaged Hermes Agent into an AWS Marketplace CloudFormation stack that deploys to a single EC2 instance in your own AWS account. IAM configured, SSL handled automatically, S3 backups running from day one, Hermes Agent v0.20.0 already installed on Ubuntu 24.04. Ready to try it? Launch the Meetrix Hermes Agent stack on AWS Marketplace.

What Is Hermes Agent?

Hermes Agent is a self-improving AI agent, which is a specific claim and not a marketing phrase. It writes its own skills from experience - ask it to do something novel a few times and it builds a reusable routine for the next time, instead of re-deriving the steps from scratch on every run.

It also remembers things. Most agent frameworks reset your context the moment a session ends. Hermes Agent keeps memory across sessions using SQLite's FTS5 full-text search combined with LLM summarization, so past conversations, decisions, and facts stay retrievable instead of vanishing when you close the terminal.

One gateway process serves Telegram, Discord, Slack, WhatsApp, and Signal at the same time, plus a terminal TUI if you'd rather talk to it directly. You're not standing up five separate bots and hoping they share state. There's a built-in cron scheduler for recurring jobs and parallel subagents for splitting bigger tasks into pieces that run concurrently instead of queuing behind each other. It isn't locked to one model vendor either - run hermes model and switch providers on the fly.

What Renting an AI Agent Actually Costs

Seat-based agent platforms bill you for headcount whether the agent does anything that day or not, and usage-metered ones charge a spread on top of raw model tokens - so the better the agent gets at your job, the worse that markup compounds every month. Running Hermes Agent on your own instance strips that layer out. You pay AWS for compute and you pay your model provider directly at their published rate, with nothing added in between. The seat-based pricing page, notably, still needs a sales call before it tells you what "enterprise" actually costs.

How Deployment Works

Setting up Hermes Agent by hand means installing Python 3.11 (most base Ubuntu images still ship 3.10 or older), Node.js, ripgrep for the skill-search functionality, ffmpeg for voice messages, a systemd unit so the gateway survives a reboot, certbot for SSL, and your own cron job to push the memory database to S3 before you can send a single message. Through the Marketplace, it's four steps:

  1. Subscribe and Launch from AWS Marketplace Open the Meetrix Hermes Agent listing, subscribe, and launch the CloudFormation stack with your chosen instance size. Hermes Agent v0.20.0 ships pre-installed on Ubuntu 24.04 and already running.
  2. IAM, SSL, and Backups Are Already Configured The stack provisions an IAM role scoped to the instance, issues an SSL certificate through Let's Encrypt via Route 53, and sets up S3 backups of the memory database, all before you've touched a terminal.
  3. Wait for Boot, Then Connect Give the instance fifteen to twenty minutes to finish starting up, then connect the gateway to Telegram, Discord, Slack, WhatsApp, or Signal, or open the terminal TUI directly over SSH.
  4. Point It at Your Model Provider Run hermes model and add your own API key for OpenAI, Anthropic, or whichever provider you use. There's no default model bundled in - you pick, and you pay that provider directly.

What Meetrix Brings to This Deployment

  • No Manual Dependency Wrangling - Python 3.11, Node, ripgrep, and ffmpeg all have to be installed and correctly versioned before Hermes Agent will even start. This stack ships with all of it already running, not a checklist to work through first.
  • IAM, SSL, and Backups Configured From the Start - Automatic Let's Encrypt certificates via Route 53, a fallback script at /root/certificate_generate_standalone.sh if your domain lives elsewhere, and S3 backups protecting the memory database from the moment the stack finishes.
  • Sized for Real Agent Workloads - t3a.small is the recommended default for a single active agent handling normal chat traffic, not a demo-sized instance that chokes the first time three platforms message it at once.
  • No Usage-Based Markup - The software fee runs from $0.001/hr on a t2.micro up to $0.019/hr on the recommended t3a.small, with 41% off on annual contracts and a 5-day free trial before anything is charged. You pay your model provider directly, at their rate.
  • People Who Actually Run This Stack - Support goes to aws@meetrix.io with a response inside 12 hours, from engineers who deploy this AMI regularly, not a support queue reading from a script.

Who Is Hermes Agent on AWS Right For?

This deployment fits anyone who wants an agent that's actually theirs, not rented by the seat. It's a strong match if you're:

  • A solo founder or small team who wants something answering questions in Slack that still remembers what it discussed last week
  • An agency running client-facing bots in Discord, where the parallel subagent model lets one instance split work across multiple client accounts instead of paying for a separate deployment per client
  • A developer who already has API keys sitting around for OpenAI or Anthropic and doesn't want to hand a third party access to them
  • A team automating on-call triage or recurring reporting through the built-in cron scheduler, across Telegram or Slack
  • Anyone who hit the usage wall on a hosted agent platform's free tier and wants to know what the same category of tool costs when nobody's marking it up
  • A compliance-conscious org that needs conversation history and model API keys to stay inside its own AWS account rather than a vendor's

Hermes Agent on AWS by Meetrix vs Alternatives

Feature Hermes Agent on AWS by Meetrix Hosted AI Agent SaaS Self-Managed Install
License MIT, fully open source Proprietary Open source, but you own every step of running it
Pricing Model Hourly instance fee from $0.001/hr, plus your own model cost, no markup Per-seat or usage markup layered on top of model cost Server cost plus your own model cost
Cross-Session Memory Yes, FTS5 search plus LLM summarization on your own instance Varies, often session-scoped Yes, if you build and maintain it yourself
Platforms From One Deployment Telegram, Discord, Slack, WhatsApp, Signal, and a terminal TUI, one gateway Usually one, more as paid add-ons Whatever you wire up yourself
Deployment Time 15-20 minutes via CloudFormation Minutes, but you're locked into the vendor Half a day or more - runtimes, gateway process, SSL, backups
Data Location Your own AWS account Vendor's infrastructure Your server
Support Meetrix engineers, response inside 12 hours Support tiers, often gated by plan Community forums only

Resources

Under the Hood

Here's what's actually running once the stack finishes deploying, in case you'd rather know that before pointing it at a model API key.

Runtime

Docker Container

Hermes Agent itself lives inside a container named hermes on the instance. sudo docker exec -it hermes hermes model is how you reach the CLI to set a provider, and the same container serves the web dashboard.

Access

nginx in Front

nginx terminates SSL and enforces HTTP Basic Auth before a request ever reaches the container, using the admin username and password you set as stack parameters (stored in /etc/nginx/.htpasswd, changeable afterward over SSH).

Interface

A Web Dashboard, Not Just a CLI

Sessions, Chat, Files, Models, Logs, Cron, and Channels tabs cover most day-to-day operation, so checking logs or adding a scheduled job doesn't require SSHing in every time.

Memory

FTS5, Not Vector Search

Conversation history persists through SQLite's FTS5 full-text search combined with LLM summarization, stored on the instance itself, not shipped off to a third-party vector database.

Data Handling

Scoped Backups and IAM

The memory database backs up automatically to the S3 bucket you designate as a stack parameter, and the instance's IAM role is limited to what the stack needs rather than broad account access.

Certificates

SSL Renews Itself

Certificates come from Let's Encrypt via Route 53 automatically. If your domain isn't delegated there, /root/certificate_generate_standalone.sh generates one manually.

One Thing Hermes Agent Won't Do For You

Memory recall here is lexical, not semantic. FTS5 is full-text search, so it matches on the words you actually used, not on meaning. Ask about "the pricing outage" three weeks after discussing it as "the billing incident," and there's a real chance it doesn't surface, because the two phrases don't share enough words for FTS5 to connect them. Vector-embedding memory systems handle that kind of paraphrase better. Hermes Agent's approach is faster and cheaper to run, and in practice most people ask about things using roughly the same words they used the first time - but it's worth knowing before an incident, not during one.

Frequently Asked Questions

What is Hermes Agent actually used for?

Team assistants inside Slack or Discord, personal automation over Telegram, scheduled jobs through the built-in cron scheduler, and general agent work where you want something that remembers what you told it last week instead of starting cold every session.

Do I need my own API key?

Yes. Hermes Agent doesn't include model access - it connects to whatever provider you configure with hermes model. That's why there's no markup on your usage: you pay OpenAI, Anthropic, or whichever provider you pick, directly, at their rate.

How much does this cost to run?

The Meetrix software fee ranges from $0.001/hr on a t2.micro up to $0.019/hr on the recommended t3a.small, plus AWS's standard EC2 charge for the instance itself and whatever your model provider bills for tokens. Commit to an annual contract and the software fee drops 41%. A 5-day free trial is available before anything is charged.

What instance size should I actually pick?

t3a.small is the recommended default and handles a single active agent across a normal amount of chat traffic. Size up if you're running heavy parallel subagent workloads or high message volume across all five platforms at once. A t2.micro is fine for testing or a low-traffic personal setup.

Does it remember things between conversations?

Yes, that's a core feature rather than an add-on. Memory persists across sessions using SQLite FTS5 full-text search combined with LLM summarization, stored on your instance and backed up to S3 automatically once the stack deploys.

How do I connect Telegram, Discord, or the others?

You register a bot with BotFather, drop the token into the Hermes Agent config, and the gateway picks it up on restart. Discord, Slack, WhatsApp, and Signal follow a similar pattern - the full walkthrough for each is in the Hermes Agent AWS Developer Guide.

Where does my data actually live?

On the EC2 instance inside your own AWS account, not on Meetrix's infrastructure and not on Nous Research's. We don't have access to your model API keys, your conversation history, or the memory database. Backups go to an S3 bucket inside your own account.

What if the automatic SSL setup fails?

The stack issues certificates automatically through Let's Encrypt via Route 53 if your domain is already delegated there. If it isn't, or the automatic path fails for any reason, run /root/certificate_generate_standalone.sh on the instance to generate the certificate manually.

Run Your Own Agent on AWS

Stop paying per seat for an agent you don't control. Deploy Hermes Agent on AWS in minutes, set up by a team that runs this stack every day.

Deploy on AWS Marketplace