Hermes Agent has been cloned more than 24 million times and, by Nous Research's estimate, drives roughly 2.5% of global AI token usage. On October 7, the company closed a $90 million Series B and announced it was turning that developer traction into a product for businesses. The round is significant. What it means for real teams is more specific.
What Hermes for Businesses actually is
Hermes for Businesses is an open-core enterprise wrapper on top of an MIT-licensed agent, not a separate product. Nous plans a paid enterprise version with private AI use, model choice by price and goal, and company-owned knowledge - the software stays free; hosting, security, management, and support are paid. That is the standard open-core playbook, and it has a solid track record (Elastic, GitLab, Grafana). The question is whether it plays out the same way for an agent that can be pulled from a public repo and wired to Slack in an afternoon.
The funding numbers give some useful context. Nous raised a $90 million Series B at a $1.5 billion valuation, led by Robot Ventures, with participation from Nvidia, Union Square Ventures, Menlo Ventures, Samsung, and 1789 Capital - bringing total funding to $158 million.
Nous was at roughly $36 million in annualized revenue by mid-September 2026 and expects to pass $100 million before the end of the year. That revenue number suggests the developer-to-paid conversion is already working at some rate - this is not a pure land-grab.
How the Slack integration works today (before you buy anything)
The current community setup is genuinely usable. The Hermes Slack integration runs over Socket Mode, so the Hermes instance can stay behind a firewall without a public inbound webhook endpoint. For a team on a private VPC or a corporate network, that matters - no public URL to expose, no reverse proxy to maintain.
The SLACK_ALLOWED_USERS setting is a security requirement; without it, Hermes denies all messages by default.
You explicitly allowlist Slack member IDs.
A single Hermes instance can behave differently in each channel, controlled through a config file.
You can bind one set of skills to #research, a different set to #sales, and keep them separate.
Beagle in action#eng-oncall, 11:42amWhat you cannot do easily with a single self-hosted gateway: give two different Slack workspaces genuinely isolated agent memory. Multiple Slack workspaces can share one gateway, but they share the same agent process and memory store unless you run separate Hermes instances. For a single-workspace team with one instance, that is fine. For an agency deploying Hermes on behalf of five clients, each in their own workspace, it means five separate installations - and five separate maintenance headaches. That is the gap Hermes for Businesses is designed to fill.
The open-source AI agent for Slack decision, made concrete
Here is the actual trade-off, put into a table:
| Dimension | Self-hosted Hermes (MIT) | Hermes for Businesses |
|---|---|---|
| License cost | $0 | Paid (pricing not yet public) |
| Slack connectivity | Socket Mode, self-configured | Managed, supported |
| Per-workspace isolation | Manual (one instance per workspace) | Built-in multi-tenant |
| Model selection | Any provider you configure | Managed, audited |
| Audit log | DIY | Included |
| Who maintains the gateway | You | Nous |
| Data leaves your infra? | No, if you run local inference | Depends on hosting tier |
Hermes for Businesses lets companies pick models by price and goal, keep their AI use private, and own their institutional knowledge - and it runs locally or in the cloud on macOS, Windows, and Linux, as well as through Slack, Telegram, and WhatsApp.
The decision tree is short. If you have one Slack workspace, an engineering team that can manage a gateway process, and a model provider already configured, the open-source path is available right now at zero cost. If you need an audit log for compliance, multi-workspace isolation, or someone to call when the gateway goes down at 2am, that is what the enterprise product is selling.
Beagle in action#product, Thursday standupWhat the $90M signals about where this goes
The $90 million funds a product built on an agent anyone can already copy - Hermes Agent is MIT-licensed - and Nous is betting that tomorrow's moats will be built around execution, trust, and extensibility rather than around the code itself.
That is an explicit bet against the idea that owning model weights is a durable advantage. Nvidia and Microsoft's M12 are on the cap table alongside Union Square Ventures. The raise moves Nous from open releases to an enterprise product while keeping the core agent open source. That structure - closed hosting on open code - is how you get both the developer adoption flywheel and a paid product. It is also what makes the threat from a competitor who forks the repo relatively blunt: they still have to build the enterprise wrapper.
The thing teams should watch is what the "company-owned knowledge" claim actually means in practice. Hermes Agent already builds and reuses skills across sessions. Ask it to research a topic on Monday, and it still remembers the findings on Friday. Whether the enterprise product ships a proper knowledge-base layer - something with searchable, permissioned context that survives agent restarts and stays scoped per department - is the detail that will separate a real enterprise product from a managed hosting tier with a better support ticket queue. Nous hasn't published specifics on that yet.
For teams already using an open source AI agent in Slack, or evaluating one, the October 7 announcement changes the calculus slightly: there is now a supported, enterprise-grade path forward from the same codebase. Whether that is better than a teammate like Beagle - which lives natively in Slack with a draft-and-approve model built in - depends on whether you want to own the infrastructure or just the work.
Open source AI agent for Slack: common questions
Can I run Hermes Agent in Slack right now without paying?
Yes. The MIT-licensed Hermes Agent connects to Slack through Socket Mode - an outbound WebSocket connection that requires no public URL. You need admin rights to create a Slack app, a configured model provider, and a server or local machine to run the gateway process. There is no cost beyond compute and storage.
What does Hermes for Businesses add over the free version?
The enterprise product targets three gaps in the DIY setup: proper multi-tenant isolation across workspaces, a managed audit log for compliance, and supported hosting so your team is not maintaining the gateway. Pricing has not been published as of October 14, 2026.
Is Hermes Agent actually private if I self-host it?
It can be. If you run a local model (via Ollama or a local provider), prompts and responses never leave your infrastructure. If you route through a cloud provider like Anthropic or OpenAI, those calls leave your network under that provider's data terms. The gateway itself is on your hardware either way.
How does Hermes Agent compare to a native Slack AI tool?
Hermes Agent is general-purpose and model-agnostic; you configure which LLM it calls. Native Slack AI tools are tightly integrated but limited to approved enterprise plans and closed models. A self-hosted Hermes instance gives you more control over the model and the data path, at the cost of operational overhead.
What is the biggest practical limitation of the current Slack integration?
Multi-workspace isolation. A single Hermes gateway can be connected to multiple Slack workspaces, but workspaces share the same agent process and memory store by default. Teams that need strict data separation between clients or departments need separate instances - which Hermes for Businesses is intended to solve.