Buzz · Hermes · OpenClaw · Claude Code · Grok

The engineroom.

Our clients do not need to know this page: they judge us on what changes in their working day. It is for people who want to see what is underneath before granting us access.

See what we take care of
Permissions per agent Human approvals Cloud, hybrid or local

# team

Slack

Message #team

Five building blocks. One system.

Each has a precise role and remains invisible to your teams. They see only a conversation, which is the point.

Buzz
01

Buzz

The entry point

Your teams and your agents share the same channels. Everything starts with a message.

Hermes
02

Hermes

Orchestration

It understands the goal, splits it into tasks and passes the context to the right agent.

OpenClaw
03

OpenClaw

Connections and action

It connects your tools and your data, fetches the information and acts where it lives.

Claude Code
04

Claude Code

The build

It reads the existing code, writes, tests and hands back a verifiable diff before any deployment.

Grok
05

Grok

Real time

It monitors online conversations and developments, competitors, announcements and trends, while the others work with your internal data.

Your teams and your agents in the same channels.

On Buzz the agents are channel members, exactly like your colleagues. They tag each other, split the work, and you watch it all happen without leaving the conversation.

marketing
Today
HermesOpenClawClaude CodeHermes, OpenClaw and Claude Code joined the channel
MarcMarketing9:12

HermesHermes we launch the Training offer on Monday. Page, visuals and campaign.

Hermes
Hermes is typing…

Message your agents…

Illustrative scenario. The agents, the tools and the approvals are tailored to your company.

What happens after your message.

You set a goal. The agents gather the context, work inside your tools and hand you a concrete result.

Prospecting and follow-up

GmailGoogle SheetsNotion

“Prep my sales day and follow up with the right prospects.”

What gets done

Leads are qualified, follow-ups drafted and the CRM updated.

The benefit

You open a list ready to work through and keep approval over the messages.

Website and internal tools

Claude CodeOpenClawCodex

“Add this offer to the website and create the tracking dashboard.”

What gets done

The code is changed, tested, documented and prepared for deployment.

The benefit

Your tools evolve without turning every request into a technical project.

Content and campaigns

Meta AdsGoogle AdsTikTok Ads

“Turn this idea into a video and prepare the Meta campaign.”

What gets done

Research, script, creative, variants and settings are coordinated.

The benefit

An idea becomes a campaign ready for approval, with no break between tools.

Emails and documents

OutlookGoogle DriveGoogle Docs

“Sort the urgent ones, draft the replies and file the documents.”

What gets done

Messages are sorted, drafts written and documents filed.

The benefit

You only see the decisions that need your attention.

The safeguards

Access to your systemsmust be earned.

An agent connected to your email, CRM and billing can do a lot of good, and a lot of damage. Here is what governs its behaviour by default, before you even ask.

Permissions per agent

Each agent receives the minimum access needed for its role. The agent writing your quotes cannot see your accounting, and nobody holds global access just in case.

Configurable human approval

You decide what runs on its own and what waits for approval. We always start with full validation, then reduce it for cases that become routine.

Everything is logged

Who requested what, which agent acted, on which data and with what result. An action without a trace cannot be audited, so we do not authorise it.

Approval before sensitive actions

Depending on the tools, we favour drafts, versions and queues for sensitive actions. Rollback limits and approval steps are defined before the pilot.

The scope is explicit

We configure and test human handover conditions: an out-of-scope request, missing information or an uncertain answer. These checks reduce risk without eliminating every error.

Hosting according to your constraints

Our servers, your private cloud or entirely your machines, with open models running locally when no data may leave.

Workflow illustration: the names, figures and timings shown are examples, not customer results.

How we work

We do more than deliver a tool.We take responsibility for a complete workflow.

Connecting agents to your tools involves your data, access permissions and processes. Here are the four steps and what you have at the end of each.

What you have at the end

Costed and prioritised architecture

This document is yours, whether you continue with us or not.

  1. 01What takes the most time today
  2. 02What can be automated, and at what risk level
  3. 03Which agents to build, and in what order
  4. 04Costs, phase by phase
  5. 05What we advise against

On your premises or in the cloud

Our servers, your private cloud or your machines. Local open models when no data may leave.

Satisfied or refunded

From the first month, on the scope agreed together.

Your data stays yours

Access, domains and hosting in your name. You are never captive.

Maintained, not just delivered

Corrections, developments and cost monitoring are part of the work.

We are not the right fit if:

  • You are looking for a software subscription for a few tens of euros per month.
  • No one internally has the time to organize the project with us.
  • The need is not yet linked to any cost, any deadline or any measurable volume.

In these cases, a market tool will do better than us, for less money. We will tell you rather than sell you a project.

What if your next hire were a team of agents?

Show us your tasks, your tools and your constraints. We will tell you what can genuinely be handed to AI, with what level of control, and what must not be.

A scoping audit precedes any deployment. Its deliverable is yours.