Skip to content

What's new in v1.3

Guided setup, start to finish. A new project starts with What is your agent for?: pick one of 16 industries (By industry) or one of the use cases (By use case), with one search across both — see the catalogue. The template you pick pre-selects skills and settings; you can change it later.

  • Editable greeting. The agent’s greeting is its own setup step, prefilled from the template. It is checked for a clear AI disclosure, and you are warned (never blocked) if it is not written in the agent’s main language, with a one-click translation.
  • Knowledge step. Add your website and documents during setup. Each website becomes exactly one knowledge source, even when several setup fields point to pages of the same site, and removing or replacing a source also clears the suggestions it produced.
  • Opening hours from your website. Opening hours found on your website or in a document are read into the opening-hours table for you to review: day ranges become one row per day, split shifts become two rows, and times of events or offers are ignored.
  • Skills and templates — a Simple project’s agent is a template plus a catalogue of skills you toggle on and off, and it can be upgraded once to an Advanced project — see Simple projects.
  • Main language and timezone. Each agent profile has a main language and a timezone (default: Central European Time, Zurich). The agent opens every call in its main language, and the voice picker lists voices that speak your languages first.
  • Call language rules. When a caller switches to another of the agent’s languages with a real sentence, the agent follows and stays in that language; a single foreign word, a name, or a “yes” does not switch the call. Fixed lines (such as the inactivity warning) are spoken in the conversation’s language. The greeting stays in the main language.
  • Barge-in. A caller interrupts the agent with one spoken word. Background noise no longer pauses the agent, and in browser sessions words spoken over the greeting are no longer lost.
  • Inactivity watchdog. After 120 seconds of silence the agent says a short warning; if the caller still says nothing, the call ends 15 seconds after the warning. This replaces the old per-flow “hang up after” setting.

See How calls behave.

  • Test calls per plan. Every project gets 20 free test calls per day; each test call has a time limit that depends on your plan. Further test calls that day count toward your plan’s minutes, and the first of them asks for confirmation. Call History hides paid test calls by default and marks them with a Test call badge when you include them.
  • Usage alerts. Owners and admins receive an email, and the project an in-app notice, at 80 % and 100 % of the included minutes. With on-demand spending on, the usage page also takes an optional warning point and stop point in overage minutes. Overage is only charged while on-demand spending is on.

The app is fully available in German, using the formal “Sie” throughout, including onboarding, the agent page, dialogs, and timezone names.

  • Flow AI edits in place. The flow AI assistant now changes the existing graph instead of redrawing it: your nodes keep their positions, and new nodes are placed next to the node they connect from. It only rebuilds the whole flow when you ask for that.
  • Authentication (per-project auth directories and the authentication node) is in preview — see Authentication node.
  • Code snippets in hooks and graph nodes — see Code snippets.
  • Platform projects in the app — creating and opening platform projects from the app. Existing platform API keys keep working — see Platform API.
  • Call forwarding and transfer to a person — see Phone numbers.
  • Per-language messages — say-node messages and consent notices written per language.

The Service API manages a Simple project’s agent (template, greeting, skills) alongside its Organization, Project, Conversation, Knowledge, Integrations, Numbers, Web Sessions, and Keys resources. The MCP server exposes 41 tools — see MCP. Every error now uses one envelope with a request id — see Errors.

With the next service release, GET /v1/me and the MCP tool get_context return the calling key’s organization, project, scope, and name, and MCP tools default projectId to the key’s own project.

The OpenAI-compatible Platform API resolves every request through a paid usage tier (standard / premium / eco), each with its own model, voice, and speech-to-text selection and concurrency behavior, on top of chat, speech, transcription, realtime, batch, and knowledge-base endpoints — see Platform API.

A first look at embedding a Nairon voice/text agent directly into your own app, without going through the Client console’s widget — design-only for now, see Voice Embed.