Comparison
One agent, several people
Running one agent for several people: one gateway with roles, one agent per person, department or client.
Posters describe four ways to serve several people. Michael Versluis says his energy advisory firm runs one Hermes gateway for the team, identifying each person by Telegram chat ID at session start and applying a role (admin, sales, marketing, calendar assistant) with fail-closed boundaries; he gives no headcount. @coldemailchris’s GTM team shares one Hermes on a VPS in Slack, with a channel per person to keep contexts apart, and says one shared Codex subscription saves about $800 a month against a seat each. For client work, an SME automation consultant gives each client its own environment and a separate container per agent, while an installer charging a monthly fee describes a client-premises hub with spokes over Tailscale. Craig Hewitt stresses building with the team, not alone. Commenters warn that once a pilot is trusted, clients ask for more agents, so plan capacity early. Evidence is thin on scale and none of it is independently verified.
Key points
- One gateway can serve several people if each user is identified and scoped by role — 1 post. (Michael Versluis · How do people run Hermes for more than one… ¶ L13)
- Keep clients, or people, in separate contexts: own environment, container or channel — 3 posts. (an SME automation consultant · Anyone else deploying Hermes into SME… ¶ L18) (@coldemailchris · Our team built an AI agent that builds our… ¶ L14) (Michael Versluis · How do people run Hermes for more than one… ¶ L19)
- Expect demand for more agents once the first one is trusted — 1 post. (an SME automation consultant · Anyone else deploying Hermes into SME… ¶ L347) (a Reddit commenter · Anyone else deploying Hermes into SME… ¶ L333)
- Share one model subscription across a team rather than buying seats — 1 post. (@coldemailchris · Our team built an AI agent that builds our… ¶ L12)
- Build agents with the team, not for it — 1 post. (Craig Hewitt · Day 4 of agentifying my business ¶ L4)
Frameworks
- Multi-person Hermes deployment options — The options witcheer offers for running Hermes for several people. (witcheer · How do people run Hermes for more than one… ¶ L2)
- One profile per person on the same box
- One gateway serving several people
- A hosted agent each on a team balance with Hermes Business
- Something else
- Role-based capability layer on one gateway — One gateway identifies each person and applies a role that limits what the agent may do for them. (Michael Versluis · How do people run Hermes for more than one… ¶ L13)
- Admin/CEO: full system access, delegation, financial oversight
- Sales/Retention: own leads, CRM deals and tasks, no financial or infrastructure access
- Marketing and Partnerships: content approval and domain workflows
- Assistant/Calendar: free/busy and scheduling only
- Role dashboards read from an Obsidian vault at session start
- Hub-and-spoke client setup — A box on the client’s premises is the hub, which the installer controls; client machines connect as spokes over Tailscale. Claude with max subscriptions handles the main work, and the Hermes API covers workflows Claude does not. (an installer charging a monthly fee · I made €2,700 last month installing Hermes… ¶ L29)
- Tailscale network
- Claude max subscriptions
- Hermes API for specific workflows
- on-prem box controlled by the installer
- own Claude-desktop-like interface with the installer’s workflows
Tactics
- Identify each user by Telegram chat ID — Everyone talks to the same Telegram bot; the agent identifies the person at session start and applies their role. (Michael Versluis · How do people run Hermes for more than one… ¶ L11)
- Load a per-user dashboard from an Obsidian vault at session start — The agent reads the user’s dashboard so it knows what is relevant without leaking cross-team context. (Michael Versluis · How do people run Hermes for more than one… ¶ L17)
- Run a separate container for each agent when a client wants more than one. (an SME automation consultant · Anyone else deploying Hermes into SME… ¶ L349)
- Give each person their own Slack channel with the agent — Separate channels keep one person’s list build out of another’s context. (@coldemailchris · Our team built an AI agent that builds our… ¶ L14)
- Run Hermes on a VPS in Slack on one shared Codex subscription — Hermes runs on a Hostinger VPS 24/7 in Slack with one Codex account underneath instead of a seat per person. (@coldemailchris · Our team built an AI agent that builds our… ¶ L12)
Stories & examples
- Energy advisory firm on one Hermes gateway: Versluis says his energy advisory firm runs one gateway for the team with roles for admin, sales, marketing and calendar assistant. Lesson: One gateway can serve several people without separate machines or profiles when role scoping is added. (Michael Versluis · How do people run Hermes for more than one… ¶ L9)
- Second-phase demand at an existing client: A department leader at one client wants an agent for email triage and project assistance and another for a research workflow; the poster is on the cusp of this phase and wonders how many agents a company could end up running. (an SME automation consultant · Anyone else deploying Hermes into SME… ¶ L347)
Numbers
- One shared subscription saves about $800 a month against a seat per person (self-reported). (@coldemailchris · Our team built an AI agent that builds our… ¶ L13)
Caveats & warnings
- witcheer asked for team size; the reply states roles but no headcount, so the scale of the setup is not evidenced. (witcheer · How do people run Hermes for more than one… ¶ L6)
- poster could not judge a commenter monitoring client machines’s setup because the poster gave no context on organisation, problem or agents per user. (an SME automation consultant · Anyone else deploying Hermes into SME… ¶ L266)
How much to trust this
- Evidence: 5 posts, 7 independent speakers (repeat posters counted once).