Conversational OS Setup

from €14,000

The base your conversational OS runs on, owned by you.

Start the conversation

Deliverable: A diagnosed operation plus a hardened, self-hosted automation base on your own server, fully documented and handed over.
For whom: Founders, COOs and IT leads who want the automation base on their own infrastructure with full data control, not on a vendor’s cloud, and set up right once.
Exit: Built to Leave. Server, instance, credentials and docs, all in your hands, nothing on ours.

Where it sits

The base (Shadowing and Design) that Conversational Operations runs on. Take it on its own if the base is all you need.

When: you want the automation base on your own infrastructure.

How it works

Running costs stay yours and stay transparent: the server at Hetzner runs about €10 to €20 per month, Claude licences €20 to €90 per person per month on Team or Max, plus API usage for the agents. Self-hosted n8n is free under fair-code and MCP is an open standard, so there is no platform fee on top. Price is reduced when your own IT takes on the server configuration.

What it is

The diagnose-and-base service Conversational Operations runs on, carved out on its own. First we get onboarded like new team members: we shadow the operation, map your data and learn how the work actually runs. Then we rent and set up your own server in a German data centre, wire your domain with automatic SSL, install the latest self-hosted n8n as the execution layer, and connect Claude over MCP so workflows get built, changed and debugged straight from a conversation.

Claude is the default; other LLMs (OpenAI, Gemini, open models) can be wired the same way, as replaceable nodes rather than a locked gateway. Everything runs on your infrastructure; your data never leaves it.

What you get

Shadowing. Onboarded like new team members: we map your data, processes and tools, and hand back a blueprint plus a prioritised build list.
Your own server, full control. German data centre, your domain with automatic SSL, all data on your server, not with third parties.
Automation platform. Latest n8n with a separate execution layer, so the UI stays fast while workflows run in the background, and an own database for workflows, credentials and run history.
AI layer. Persistent agent memory, cleanly separated per user and channel, guardrails against runaway loops and agent failures, and MCP so workflows are built and debugged through Claude. Model-agnostic.
Security. SSH hardening with key-only access and no password login, automatic blocking of suspicious access through Fail2ban, credentials encrypted and only ever on your server.
Maintenance. Backup before every update and major change with retention logic, automatic cleanup of old history and file remains, self-monitoring that restarts hung processes, and a defined update process.
Skills and connectors. All n8n Skills plus the Human Communication Skill installed, API connections to your tools in n8n, and the connectors set up in Claude.
Governance, optional. Access tiers, for example one n8n-MCP everyone can reach and a restricted one for a defined circle.
Handover. End-to-end functional test of all core functions, and the full setup documented as a Claude project so your team adjusts and debugs the server directly.

How it ends

Built to Leave. Server, instance, access, credentials and docs are all in your hands. Self-hosted n8n is free under fair-code and MCP is an open standard, so you pay server and usage only, and you can run, extend and debug it without us.

Where this sits in the four phases.

Every service runs through the same method: shadow the live operation, design the architecture, operate with your team, then hand over.

Phase 0

Shadowing Blueprint

Diagnose in your live operation. Decision logic, data flows, role boundaries. We watch how the work actually gets decided before we touch anything.

Phase 1 This service

Design Architecture

Channel logic, audience model, KPI framework, MarTech, workflow. The operating system on paper, agreed with your leadership before we build.

Phase 2

Operate Running system

Operate with your team. Forecasting, media buying, automation. Enablement runs in parallel so ownership moves to your people while the system runs.

Phase 3

Control Handover

Attribution, incrementality, business impact. Then the handover cut: documentation, access, and usually the full-time leader we helped you hire.

Runs well with

Start the conversation on Conversational OS Setup

No cart, no checkout. You tell us the situation, an operator reads it, and we come back with scope and price.