Hermes by Polar
Open-source autonomous agents

Autonomous agents that
run your ecommerce.

Hermes is the open-source AI agent from Nous Research, and Polar makes it perform on your ecommerce data. You get autonomous agents that execute real tasks across your connected tools, from Amazon Seller Central to Meta Ads, and read, report and act wherever your team works. What makes them perform is Polar's data foundation: a governed semantic layer and 45+ native integrations feed every agent clean, attributed data, so they run on another level and never hallucinate your numbers. Run it yourself, or have the Polar team deploy yours white-glove.

@ ⏎
Deploy Hermes with Polar See how it works
What is Hermes?

Hermes is the open-source AI agent from Nous Research. Polar makes it perform on your ecommerce data.

Instead of a chatbot you query, Hermes gives you agents that execute real work across your connected tools, from Amazon Seller Central to Meta Ads and Klaviyo, on a schedule and wherever your team works. They monitor, brief, investigate and propose the next move, and they act only when you approve. What makes them actually perform is Polar's data foundation: a governed semantic layer and 45+ native integrations feed every agent clean, attributed numbers, so they run on another level and never hallucinate your metrics. Hermes is open source, so you can run it yourself, and because it is powerful we recommend the Polar team deploy and tune yours white-glove.

Open source

Hermes is MIT-licensed by Nous Research. Run it yourself, or have Polar deploy it for you on your governed data.

Agents that act, not just answer

They watch your business, report on schedule, and propose the move, then execute across your tools once you approve.

Runs on governed data

Every number comes from Polar's semantic layer across 45+ tools, so blended ROAS, true CAC and margin match your books.

Deploy Hermes with Polar Explore the open source
Setup Guide

Deploy autonomous agents, step by step.

Connect your data in Polar, we deploy your agents wired to your tools, then let them run. The data foundation is what makes them perform. Hermes is open source, and we recommend a white-glove deployment by the Polar team.
01 · What you need
Polar Analytics

Polar, your data foundation

A governed semantic layer plus 45+ native integrations. It is the source of truth every agent runs on, so they all act on the same correct, attributed numbers.

Connect your data →

Your tools connected

Amazon Seller Central, Meta, Google, TikTok, Klaviyo, Shopify and more. Agents act where your business runs, and report wherever your team works: Slack, Telegram, Discord.

See the integrations →
Polar team

The Polar team to deploy it Recommended

Hermes is open source, so you can run it yourself. Because it is powerful and only as good as the data under it, we highly recommend the Polar team stand up and tune your agents white-glove, wired to your data and your tools, in a few days.

Deploy with Polar →
02 · Three phases to agents that run themselves
1

Connect your data in Polar ~5 minutes

Build the data foundation your agents run on.
How to
1
In Polar, connect your stack through 45+ native integrations: Shopify, your ad platforms, Klaviyo, Amazon, GA4 and more. Polar's semantic layer turns raw exports into governed metrics: blended ROAS, true CAC, contribution margin, defined once and consistent everywhere.
2
Make sure your data sources are connected, then wait for the first sync. The initial historical sync can take up to 24 hours for a new Polar account.
Why it matters: an agent is only as good as the data under it. Polar's foundation means the numbers your agents act on match your bank account, not a cookie estimate or an LLM's guess.
Connect your data in Polar
2

We deploy your autonomous agents White-glove · recommended

Dedicated agents, wired to your data and your tools.
How to
1
The Polar team stands up your dedicated Hermes agents, connected to your Polar data foundation and your platform connectors (Amazon, Meta, Google, Klaviyo and more).
2
Scope one agent per lane, each in its own Slack channel: media buying, inventory, retention. Each stays focused, only sees what it should, and works where you do: Slack, Telegram, Discord, your call. One brand runs a single Hermes across four channels and brought a whole agency in-house.
3
Onboard each agent like a junior hire. The lane's owner, its channel champion, talks to it, gives it context up front and checks its work. It remembers everything you tell it, so it gets sharper every day and earns more autonomy as it proves itself.
Open source, better with Polar: you can run Hermes yourself, but it is only as good as the data under it. We highly recommend the Polar team deploy and tune your agents white-glove, so they get sharper fast and act on numbers you can trust.
Deploy with Polar
3

They run the work autonomously the payoff

Hand off the task once. The agent executes it across your tools, again and again.
What this unlocks

One instruction becomes an agent that acts.

Describe the job in plain English and Hermes turns it into an autonomous agent: a media buyer, an Amazon operator, an inventory planner. It runs on a schedule, pulls live data from across your connected tools, drafts the work, proposes the changes it would make, and reports the impact. For anything with numbers it runs a panel of sub-agents that check each other first, so it acts on a figure you can trust, not a confident guess.

One merchant brought their Amazon ad agency in-house, running on Polar and Hermes. Agents propose the moves, a human approves, and the impact shows up in the numbers.

A few of the agents you can put to work:

AGENTMedia Buyer: flags fatigued creative and where to shift budget across Meta, Google and TikTok.
AGENTAmazon Operator: tracks bids and search terms and recommends moves to hit your TACOS.
AGENTInventory Planner: flags reorders and dead stock from live sell-through.
AGENTRetention: drafts and triggers win-back flows in Klaviyo.
AGENTCreative-LP Auditor: finds ads pointing to the wrong landing page.
AGENTDaily Brief: posts a governed P&L to Slack, Telegram or Discord.
Plug a generic agent into raw data and it will confidently make numbers up. Hermes agents run on Polar's governed semantic layer across 45+ connected tools, so they act on the truth, and they will even tell you when a metric is not safe to trust yet, like when a data source is still syncing.
See the agents you can run
03 · Prefer to self-host? · open source

Hermes is the open-source agent from Nous Research, so your team can run it yourselves. You bring the data foundation with the Polar MCP, so the agent acts on governed, attributed numbers instead of guessing.

How to
1
Install Hermes from Nous Research: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash (macOS/Linux) or download the desktop app. A Nous Portal account covers the models.
2
Connect the Polar MCP so the agent runs on your governed data (Shopify, Meta, Google, Klaviyo, all clean and attributed). Follow the Polar MCP setup guide.
3
Connect Slack (Hermes ships adapters for Slack, Telegram, Discord and more) so the agent lives where your team works.
4
Brief an agent and put it on a schedule. Use any job from the library below as the standing instruction: it proposes, you approve, it acts.
Powerful, so we still recommend white-glove: an agent is only as good as the data under it. If you would rather not run it yourself, the Polar team deploys and tunes your Hermes on your governed data.
Get started

Autonomous agents, running on a data foundation that's actually right.

Connect your data in Polar, we deploy your agents, then let them run. Polar's foundation feeds every one clean, attributed data, so they act and perform instead of hallucinate.

Get started by booking a demo See the agents ↗
What They Do

Agents that run the work, not just the reports.

Each one is an autonomous agent that executes real tasks across your connected tools, on a schedule, and reports wherever your team works. These are a few your team can run.
Media Buyer
Spends smarter, every day
Watches ROAS and CAC, flags fatigued creative, and recommends where to shift budget toward what converts. It surfaces the moves; you make the call in-platform.
Runs onMetaGoogleTikTok
Amazon Operator
Runs your marketplace
Tracks bids, search terms and TACOS, flags wasted spend and hero ASINs, and recommends the moves to make. An in-house analyst, minus the agency.
Runs onSeller CentralAmazon Ads
Inventory Planner
Never stocks out, never overstocks
Tracks sell-through and days of cover, flags what to reorder before a stockout, and surfaces dead stock tying up cash.
Runs onShopifyPolar
Retention
Keeps customers coming back
Watches cohort repeat rates, drafts and triggers win-back and post-purchase flows, and flags churn risk before it shows up in revenue.
Runs onKlaviyoShopify
Creative-LP Auditor
Sends every click to the right page
Finds ads pointing to the wrong landing page, ranks the waste by spend, and proposes the swap with the expected CVR lift.
Runs onMetaPolar Pixel
Daily Brief
Reports the business, on schedule
Posts a governed daily or weekly P&L and KPI brief, with the wins, risks and actions, to wherever your team already works.
Reports toSlackTelegramDiscord
The data foundation under every agent

This is why they work, and others hallucinate.

Most agents run on raw exports and guess your metric definitions. Hermes agents run on Polar's governed semantic layer, pulling from 45+ native integrations. Same agent, real data: it operates on another level, and the numbers match your books.

45+
native data
integrations
0
hallucinated
numbers
Agent Library

Agents you can put to work, right from Slack

Each card is a standing agent you hand Hermes once. It monitors, briefs, investigates and proposes moves on your governed Polar data, and only acts when you approve.
Mode
Audience
58 of 58
Three Pieces, One Flow

Your tools. Polar's data. Agents that act.

Polar pulls from 45+ integrations into one governed semantic layer. Hermes agents run on that foundation and execute across your stack, so they act on the truth, your definitions of ROAS, CAC and margin, not a guess.
01 · Connect

Your stack

45+ tools, one source of truth

02 · Ground

Polar

Governed semantic layer, no hallucination

03 · Act

Hermes agents

Run tasks autonomously

The Foundation

Other agents hallucinate. Yours run on a real foundation.

Most agents run on raw exports and guess your metric definitions. Hermes agents run on Polar's governed semantic layer across 45+ connected tools, so blended ROAS, true CAC and margin match your books. Same agent, real data: it operates on another level.
Get your Hermes agents
Open source · deployed by Polar · runs on your tools

Agents that run the work, on data that's right.

Hermes gives you autonomous agents that act across your tools. Polar gives them the data foundation, a governed semantic layer and 45+ integrations, that makes them perform instead of hallucinate. We deploy and tune them with you.

Deploy Hermes with Polar Book a demo
Phase 1 ~5 minConnect your data in Polar45+ integrations into one governed semantic layer, the foundation your agents run on.
Phase 2 White-gloveWe deploy your autonomous agentsDedicated agents wired to your data and your tools. You scope them per lane and onboard them like junior hires.
Phase 3 the payoffThey run the work autonomouslyAgents execute tasks across your stack, propose changes you approve, and report wherever you work.