Agent 2.0 (beta) is now available

A more flexible way to build AI agents is here. Agent 2.0 gives you simpler setup, more control over handovers, and stronger AI capabilities—while Agent 1.0 remains fully available.

What's new

More flexible agent creation

Agents can be created and saved without a connected channel or knowledge base
Agents can work purely with AI actions, no channel required

Simplified prompting model

Prompts are now focused on three core elements:

Tone of voice
Company description
Additional instructions

Any extensive or specific guidance can be added to additional instructions, and the AI will follow it.

Improved escalation and handover control

More clarity and flexibility in when and how conversations are handed over to human agents.

Powered by GPT-5 by default

Significantly more capable than the default model in Agent 1.0.

Why?

Agent 2.0 gives teams the flexibility to build and test AI agents faster, with clearer control over when the AI should hand over to humans. The simplified prompting model makes it easier to guide AI behavior without complex configurations, while GPT-5 delivers noticeably better responses across languages and use cases.

January 7