Hermes: TillClawn 🤡 Got a Brother – and It's Not a Parcel Service

Upfront

As of today, next to TillClawn 🤡 there's a second building block in our stack: Hermes. And since the question is going to come up: no, it's not a parcel service. No neckties either. And it isn't a second product we're trying to sell you – Hermes is a framework, a standalone layer inside the OpenClaw ecosystem.

Hermes is the Greek messenger god – and that's pretty much what the layer does. It carries messages between humans, agents, and channels. It builds no skill of its own, orchestrates no enterprise workflows. It just makes sure the right message lands in the right place in the right shape.

Why a second layer?

Six months of TillClawn 🤡 in production taught us one thing we had underestimated:

Per-customer value doesn't come from the skills. It comes from an agent answering in a Slack call, from a voicemail landing as a structured task, from a customer asking on WhatsApp and the answer being in the CRM before sales even reads it.

So far, that was always part of TillClawn 🤡. And it was a mistake. Because:

• Projects that only needed the messenger function had to drag along the full enterprise stack.
• Projects that needed the enterprise stack got the voice pipeline shipped along, even though they never switched it on.
• Our roadmap pulled in two directions at once – deep into skills and wide into channels.

So we cut. TillClawn 🤡 stays what it is. Hermes becomes the standalone framework layer for everything that carries messages.

What Hermes is – and isn't

Important for anyone reflexively looking for the comparison:

• Hermes is not an OpenClaw killer. It sits on an OpenClaw or TillClawn 🤡 backend. Without an agent layer it has nothing to say.
• Hermes is not a voicebot builder. If you want to build a bot that takes bookings and nothing else – use Retell or Vapi. Hermes is overkill for that.
• Hermes is not a replacement for TillClawn 🤡. It's the framework you put in front of or next to it so humans and agents can meet in the first place.

The reason for the split (technical)

Until June the voice pipeline ran as an optional layer inside TillClawn 🤡. That created two problems:

1. Latency budget was shared. A skill call against SAP is allowed 4 seconds. A voice turn is allowed 400 ms. Building both in the same process model doesn't hold up cleanly.
2. Release cadences were incompatible. Voice models (Gemini Live, ElevenLabs, OpenAI Realtime) spin weekly. Enterprise skills need monthly, documented releases. Two cadences, one repo – doesn't hold.

The split was overdue. We should have done it three months ago.

Where Hermes goes first

Three scenarios we start with:

• Inbound voice → task. Customer calls, Hermes picks up, clarifies intent, opens a monday item with a structured summary. Agent response optional.
• Meeting attendance. Hermes joins Google Meet or Teams as a named participant, transcribes, asks back when needed, files follow-ups in the CRM.
• Omnichannel routing. WhatsApp, email, Slack, SMS in one session. Hermes normalizes the message, hands it to the agent layer (TillClawn 🤡 or directly to a model), and formats the response back into the channel.

What Hermes explicitly does not do: make its own decisions. It's messenger, not advisor.

What this means for TillClawn 🤡 projects

Short: nothing you don't want.

• If you run TillClawn 🤡 today with the voice pipeline, you'll migrate to the Hermes layer in July – same functionality, cleanly separated building block.
• New TillClawn 🤡 setups no longer get Hermes bundled by default. If you want voice, you pull the layer in explicitly.
• Internal complexity in TillClawn 🤡 drops accordingly – less overhead in the enterprise layer, more focus on skills and compliance.

If you only ever wanted the voice pipeline but had to drag the full enterprise layer along to get it: from now on, Hermes ships solo.

Why the name

Because "TillClawn 🤡 Voice" was boring. And because we have a rule: if a building block has its own roadmap, its own cadence, and its own audience, it gets its own name.

Hermes is the fastest of the Olympians, crosses borders between worlds, and delivers messages. No other name described what the thing does as precisely.

And yes, "isn't that the courier?" will be the most frequent question of the next three months. We can live with that.

Verdict

Two building blocks, cleanly split:

• TillClawn 🤡 – the enterprise layer on OpenClaw. Skills, gateways, compliance, audit. For ops, compliance, CIO.
• Hermes – the messenger layer. Voice, meeting attendance, omnichannel routing. For sales, support, marketing.

No merge, no universal tool. The lesson from six months of OpenClaw in production was clear: specialization beats suite. We're building accordingly.

---

Curious about Hermes? Talk to us – first pilots run in July.

More context: TillClawn 🤡 in Production · OpenClaw Audit 2026