Braid

AI agent

Shape how it sounds, and where it stops

The agent has a name, a tone and languages you choose. More importantly it has edges — the things it will not say, and the moment it stops trying and fetches a person.

  • A persona you set

    Name, tone, and the languages it replies in — or let it answer in whatever language the customer wrote in.

  • Guardrails

    Topics it must refuse and instructions it always follows, written in plain language rather than configuration.

  • Handoff rules

    Hand over when the customer asks for a person, when they sound frustrated, or when the answer is not in your sources. Each is a switch.

  • Follow-ups

    When a conversation is waiting on the customer, the agent can check back once — in their language — rather than letting it go stale.

  • Draft and publish

    Persona and guardrail edits stage as a draft. Customers keep seeing the live version until you publish, so nothing changes underneath a running conversation.

  • Pause it

    One control stops the agent answering on every channel, and new conversations wait for a person.

Set the edges before you set it loose

Test the persona in a sandbox against your real knowledge, then publish when it sounds like you.