Skip to content
openclaw-os
Config

Config & profiles per role.

A single openclaw.json doesn't scale for a productive team. We split config into profiles per role, channel and tenant — cleanly versioned.

The problem

Why this matters right now.

OpenClaw loads a global openclaw.json from your home folder. Once you have several roles, channels or tenants, the file becomes messy, secrets leak in plaintext, and nobody knows which config is currently active.

Our approach

Here's how we do it.

We split config into profiles (sales, support, operations …), keep secrets out via macOS Keychain or systemd credentials, document model routing per provider and version everything in a private Git repo.

What's included

The full package.

PR

Role profiles

Sales, marketing, operations, IT — each profile has its own skills, channels, allowlists and models.

MR

Model routing

Fast models for triage (Haiku, gpt-4o-mini), strong ones for drafts (Sonnet, gpt-5), local for sensitive data (Llama 3.3 via Ollama).

SC

Secrets

API keys not in plaintext: macOS Keychain, systemd credentials, 1Password CLI or Vault — referenced cleanly.

VG

Versioned

Config in a private Git repo, with pull requests for every change — auditable, rollback-able.

LM

Limits

Token budgets, rate limits per skill and channel, automatic escalation on overrun.

EX

Override path

Individual users can set override defaults in their own sub-profiles without breaking the team standard.

How it works

From first call to a productive OpenClaw workflow.

  1. 01

    Analyse

    30-min check + process map. We pinpoint the 3 workflows where OpenClaw saves time fastest, and which channels are mandatory.

  2. 02

    Setup

    Daemon, channels, skills, MCP, allowlists, requireMention, Tailscale and the live dashboard configured cleanly and live.

  3. 03

    Training

    Your team works on real tasks: WhatsApp inbox triage, weekly Slack-DM reports, voice-memo-to-CRM, skill maintenance.

  4. 04

    Operations

    calver updates, skill extensions, channel care, security reviews, backups and an emergency off-switch — so OpenClaw doesn't fade out.

Packages

Three entry points. One outcome: OpenClaw that works.

All packages & add-ons →

Starter

For small teams and 2–3 workflows on one channel

€1,850 one-time

+ from €180 / month maintenance & care

  • OpenClaw daemon installed, hardened, monitorable (launchd/systemd)
  • 1 messaging channel (WhatsApp, Telegram, Slack or iMessage) cleanly paired
  • 2–3 productive skills (inbox triage, research or reporting)
  • 2 MCP tool integrations (e.g. Drive, Notion, Slack, CRM, calendar)
  • Allowlists, requireMention, GDPR defaults and backup plan
  • 2-hour intensive training + 30 days of care
Book a check

Enterprise

For sensitive data, compliance and fleet rollout

from €14,900 one-time

+ from €1,290 / month maintenance & care

  • Enterprise setup with central config, fleet auth and audit trail
  • Custom MCP server or internal tool adapters incl. code review
  • 10+ skills, departmental playbooks, train-the-trainer programme
  • Voice, canvas, Tailscale, backup and SIEM integration
  • Governance documentation for ISMS, GDPR and ISO 27001
  • Monthly security and optimisation reviews
Book a check
Frequent questions

Still open questions?

Write us at hello@openclaw-os.com or book a call directly. We'll take the time.

Does OpenClaw switch profiles automatically?
Not out of the box. We add a wrapper script and/or `OPENCLAW_PROFILE` env vars so profiles activate per channel or role.
Can I run multiple LLM providers?
Yes. OpenClaw supports OpenAI, Anthropic, Mistral, local Ollama models and more. We route tasks dynamically by sensitivity, latency and cost budget.
What if my team starts hand-rolling configs?
As long as the team profile is active, team defaults win. Members can add skills only in a clearly scoped sub-profile.
What does a complete OpenClaw setup cost?
Starter starts at €1,850, Business at €5,900 and Enterprise from €14,900. Ongoing maintenance starts at €180/month. LLM costs (OpenAI, Anthropic, local models) run separately.
Is OpenClaw only useful for engineers?
No. That's the whole point of openclaw-os.com: OpenClaw becomes a work assistant for inbox, offers, reports, meetings, research, knowledge and routines — over the channels your team already uses (WhatsApp, Slack, Telegram, iMessage, voice).
Which channels make sense?
We usually start with the channel your team already lives in: WhatsApp and Telegram for external comms, Slack/Teams internally, iMessage on macOS, Discord for communities. The multi-channel strategy follows reality, not technology.
Next step

Book a call.
30 minutes that pay off.

Pick a slot — we confirm automatically and send you the Google Meet link.

Intro call · 30 min · free

Pick a slot that works.

No commitment. We review your OpenClaw use cases, surface three productive levers, and send a short recap afterwards.

Booking loads only after your click

Clicking the button loads the Cal.com calendar. Data may be transmitted to Cal.com at that point.

Open on Cal.com

The site works fine without Cal.com. Booking is a separate, voluntary feature.

Powered by Cal.com · The calendar loads only after your action. Trouble? Write to hello@openclaw-os.com.

This site only uses technically necessary features. Analytics loads only after consent. Cal.com booking loads only when you actively open it.