Your assistant,
on your own server.
The assistant on every InAgentic server is Hermes Agent, the open-source agent built by Nous Research (MIT licence). You chat with it in plain English; it answers, uses tools, remembers context and runs scheduled jobs, and it can hand work to the four other agents on your server.
Runs on an open-source model on your server by default. Your own OpenRouter or Anthropic key is optional.
- Made by
- Nous Research
- Licence
- MIT, open source
- Model
- Qwen3 30B (local)
- Context
- 64k tokens
- Tools
- Your 5 agents
- Commands
- In its own container
Chat assistants forget, and live on someone else’s cloud.
Most assistants start from zero every conversation, can’t act on your own systems, and send everything to a provider you don’t control.
- Context is lost between conversations
- Recurring jobs still need a person to start them
- Each tool has its own assistant and login
- Your conversations sit on a third-party cloud
Chat, tools, memory and schedules.
Capabilities described by Nous Research for Hermes Agent, as installed on InAgentic servers.
Chat
- Answers in plain English
- Streams replies as it works
- Shows which tool it is using
- Takes file attachments
Memory
- Persistent memory across conversations
- Recalls earlier sessions by search
- Builds a picture of how you work
Skills
- Creates skills from experience
- Improves them as it uses them
- Skills follow the open agentskills.io standard
Scheduled jobs
- Reports, backups and briefings
- Set up in natural language
- Run unattended on the server
Your other agents
- Kai: KPIs and targets
- Coding agent: tickets to pull requests
- Max: content drafts
- Nora: call notes and roleplay
Tools and sub-agents
- Connects to MCP servers
- Runs commands in its own container
- Isolated sub-agents for bigger tasks
Ask once. Hermes picks the tools.
Type a request
In the chat on your server, for example “build ticket ABC-12” or “what did I agree with Peter?”.
Hermes chooses tools
It decides which tools or agents to use and shows its progress as it goes.
You stay in control
Actions that need your OK appear as approvals in the chat.
It learns
Useful facts and skills are kept for next time, on your server.
The first reply in a conversation takes longer on the local model (expect 20–60 seconds); after that answers stream at roughly 10–20 words a second. For faster answers, switch Hermes to OpenRouter with your own key.
Wherever you work on your server.
Web chat
On your own address (yourname.inagentic.ai): streaming replies, tool progress, approvals and attachments.
Admin
A Hermes page with its status and skills, and Model & tools to change the model it uses.
Your own keys, optional
Add an OpenRouter or Anthropic key on Settings & keys to use a frontier model instead of the local one.
Audit trail
What the agents did, in one place in the admin.
Where your conversations go.
The model
By default Hermes uses an open-source model (Qwen3 30B) running on your server, so prompts aren’t sent to an AI provider. If you add an OpenRouter or Anthropic key and switch to it, prompts go to that provider.
What it keeps
Conversations, memories and skills are stored by Hermes on your server, and included in the nightly backup.
Commands
Commands Hermes runs execute inside its own container, which can’t reach Docker or the other containers’ files.
Access
Hermes isn’t on the internet: only the web app is, behind your admin login. INAGENTIC LTD runs and maintains the server, so our operations team has technical access for updates, backups and support.
Not there yet.
Planned, partly built or not yet set up on InAgentic servers. We'll say so here until it works.
Messaging apps
ComingHermes itself supports Telegram, Discord, Slack, WhatsApp, Signal, email and more. Connecting them on InAgentic servers isn’t set up yet.
Web search and browsing
ComingHermes lists web search, browser automation and image generation; on InAgentic servers these aren’t configured yet.
Voice
ComingHermes has a voice mode for some platforms; not available in the InAgentic web chat yet.
Mobile app
ComingNous Research describes a Hermes Agent mobile app as in beta.
Questions people ask.
Is this the same Hermes as Nous Research’s?
Yes: Hermes Agent, the open-source agent by Nous Research, installed and configured on your server by InAgentic, with your five agents’ tools connected.
Do I need an API key?
No. It works with the open-source model on your server. Adding an OpenRouter or Anthropic key is optional, for faster or stronger answers.
Why is the first answer slow?
The local model runs on the server’s CPU and Hermes sends a long setup prompt first. Expect 20–60 seconds for the first reply in a conversation.
Is it Hermes Cloud?
No. Hermes Cloud is Nous Research’s own hosting. InAgentic runs Hermes on your own dedicated server or Mac mini.
Who do I contact?
Your team's administrator, or INAGENTIC LTD at axel@inagentic.ai.