Home/Hermes AI Agent
Hermes Agent · enterprise installation

Hermes Agent: secure installationfor your business.

Scroll installs Hermes Agent on a VPS or dedicated infrastructure, connects it only to the necessary tools, and defines its permissions, costs, and sensitive actions.

Hermes Agentdedicated instance
Private channelSlack · Telegram
minimal permissions
Hermes AgentDocker · VPS
Memory & skillsdedicated profile
AI modelcloud · local
Business toolsMCP · API · n8n
Approvalshuman in the loop
Supervisionlogs · costs · alerts
Actions loggedAccess revocable
01 — Recommended use cases

An agent for a specific task,
not to « automate everything ».

We start with a repetitive, verifiable, and reversible task. The agent’s permissions expand only once results and safeguards are validated.

01

Daily monitoring & briefing

Monitor defined sources, synthesize useful information, and deliver a sourced briefing in Slack or Telegram at a set time.

02

Executive or project co-pilot

Prepare meetings, consolidate decisions, and highlight next steps without losing the context built between sessions.

03

Sales preparation

Research an account, summarize useful signals, and draft a follow-up. The salesperson retains approval before any send-out.

04

Internal documentation assistant

Query a documentation base or internal tools with limited access, cite sources, and flag when an answer requires verification.

05

Technical triage in an isolated environment

Analyze issues, logs, or CI results, then prepare a diagnosis and corrective actions without direct access to production.

06

Research & draft production

Turn intelligence into plans, notes, reports, or working content. Publishing and external actions remain subject to human validation.

02 — Security by default

A useful agent,
without handing it the keys to the business.

Hermes can read files, use a browser, call APIs, and execute tools. We therefore treat isolation, secrets, and validations as core product features—not a last-minute add-on.

Dedicated and isolated instance

Hermes runs in a dedicated environment under a non-privileged user. Accessible volumes, network, and resources are limited to actual needs.

Minimal secrets & permissions

Each integration uses a service account with restricted rights. Keys are separated, revocable, and never placed in prompts or skills.

Human in the loop

Read-only during the pilot, with capabilities opened progressively. Any sensitive send, publish, or modification requires explicit approval.

Supervision & reversibility

Logs, cost tracking, alerts, backups, and tested restoration. Configuration, skills, and accounts are transferable to your team.

They trusted usView our client cases
Imalize
Sistr
Art Explora
Perfway
Hexa
Bellman
Secure pilot offer

One use case, one instance,
defined success criteria.

The pilot includes a dedicated instance, a model provider, a channel, a business skill, backups and monitoring, five test scenarios, and training. Infrastructure, APIs, and third-party subscriptions remain under the client’s name and are billed separately.

Frame my pilot
3 900 €ex VAT, starting from, for a pilot with a defined scope
1priority business use case, measured before and after
14 daysof hypercare after pilot deployment
03 — Architecture

Hermes at the core,
limited permissions around it.

The model, channels, and connectors are not selected by default. Each component is chosen based on the use case, data, cost, and risk level.

LayerPossible choiceRole and safeguards
AgentHermes AgentMemory, skills, tools, scheduled tasks, and delegation
AI ModelMistral, OpenAI, Claude, localModel selected based on quality, confidentiality, cost, and latency
Private channelSlack, Telegram, DiscordAccess via allowlist and dedicated accounts
IntegrationsMCP, API, n8nTargeted access to business documents and tools
ExecutionDocker, dedicated VMProcess isolation, resource limits, and minimal volumes
SecurityApprovals, secrets vaultMinimal permissions, rotation, and human validation
OperationsLogs, alerts, budgetsError tracking, costs, scheduled tasks, and dependencies
ContinuityBackup, restore testingControlled updates, rollback, and handover to your team
Agent coreBuilding blocks and safeguards around it
04 — Deployment

Hermes Agent on VPS or on-premise —
we choose based on your data.

Installing Hermes Agent on a VPS does not automatically mean all data stays in-house: the model provider and connected tools are part of the flow. We precisely document each destination.

Dedicated VPS + cloud model

Recommended

Hermes runs in isolation on a dedicated instance, while inference is handled by a chosen model provider configured with client-owned accounts. Third-party flows are documented.

Recommended when

You need an agent available continuously, without purchasing GPUs, with controlled budget and permissions.

Local infrastructure and model

Custom

Hermes and the model run within your environment. This architecture reduces flows to external providers but requires more hardware, operations, and performance testing.

Recommended when

Data is sensitive, usage is stable, and the organization has the appropriate hardware or infrastructure.

05 — FAQ

Hermes Agent for businesses

Installation, costs, local models, security, and compliance: the key questions to address before entrusting tools and data to an agent.

Hermes Agent is an open-source AI agent developed by Nous Research. It combines memory, skills, tools, scheduled tasks, delegation, and integrations. It can use different model providers or a local model.

Installing the software is the shortest part. The service covers use case selection, architecture, permissions, accounts, integrations, skills, testing, training, backups, and maintenance.

Hermes is free under the MIT license. You must add infrastructure, model provider, and any connected tools. We estimate these costs during scoping and set up caps and alerts.

Yes, via Ollama, vLLM, or another compatible endpoint. However, an agentic model requires a large context window and sufficient memory. We validate hardware and performance before recommending this approach.

Yes. We configure the channel, service accounts, allowlists, and necessary roles. For professional use, we avoid unofficial bridges when service continuity is critical.

Not automatically. Compliance depends on the data, model provider, connected tools, retention periods, and any potential transfers. We document the flows and technical measures; the client retains responsibility for the legal validation of their use.

Hermes converses, searches, reasons, and adapts to variable requests. n8n executes explicit, deterministic workflows. We often combine them: Hermes prepares or selects, while n8n executes sensitive operations within a controlled framework.

We do not automate without human validation payments, deletions, publications, or high-impact decisions. We also refuse public chatbots or shared instances between clients without a specific isolation architecture.
Get started

A first use case to entrust to a Hermes agent?

We start with a frequent, measurable, and reversible task. If Hermes isn’t the right tool, the scoping will reveal it before deployment.

Contact details
contact@agence-scroll.com
+33 6 48 03 90 27
20 Rue des Taillandiers
75011 Paris
Response within 24 business hours.