Hermes runs on Nous Research's open models, which some teams prefer for anything that speaks to customers in their own name. The thinking this agent needs is everyday language work: decode a half typed message, decide whether the playbook covers it, write one plain sentence back. Anything outside the playbook goes to the person you named instead of being improvised.
Run a staged conversation — no account needed. The agent handles it for real while a simulated world answers its tool calls; nothing touches real accounts, and nothing is actually sent.
Ana Sofia Guzman+52 55 1234 8890
hola, cn i chnge the delivery to fri? order 88214, nobodys home thurs. thx
Using this template drops you into a guided setup. It asks exactly this, nothing else:
Connect Twilio
One sign-in. The agent acts through your account, scoped to what this template uses.
Business name
The name your customers know you by, written exactly as you would put it in a text message.
Support playbook
Everything the agent is allowed to answer - your hours, delivery and returns terms, the common fixes, and the exact wording for anything sensitive. It only ever answers from this.
Escalation contact
Who takes over when the agent cannot help - the name and number or team, and when they are reachable, written so the agent can say it in one short line.
Runs on Hermes
Preselected for this page — connect your Hermes account during setup, or switch to NoClick's built-in models with one click.
Watch it handle a test run
A staged conversation against a simulated world — then it’s live.
The model writing to your customers is one you can name and read about, which matters to teams that have a policy about it.
Working out that a message reading cn i chnge the delivery to fri is a reschedule request is ordinary comprehension. That is the load this harness carries here.
The containment lives in the instructions and the tools rather than in the model: it answers only from your playbook, sends no link you did not supply, and hands over on money, account access or anger. Rehearse it on the three staged texts, particularly the double charge one, because tone problems surface there first. A reply that reads wrong is usually a missing playbook line rather than the model.
Free to start. Guided setup, a test run against staged conversations, and it's live.