Skip to main content
VARTA

Can I create and edit VARTA agents through the API?

No — not yet. /v1 is the runtime surface: it loads an agent you've already built, holds a conversation with it turn by turn, places and monitors telephone calls, and registers the telephony you dial through. It does not author agents.

What's absent

There is no /v1 way to:

  • Create an agent.
  • Build or edit its conversation flow — steps, branches, the sequence a call moves through.
  • Define its rules — what counts as a match, what triggers a jump to a different step.
  • Configure its routing — where a given intent or slot value sends the conversation next.
  • Set its tone — how the agent is meant to sound and respond in character.
  • Author its reflex phrases — the fixed responses it reaches for on specific triggers.

All of that happens in the VARTA studio today. If you've read the Quickstart, the agent id you loaded in step one — the thing /v1 treats as an opaque, already-built unit — is exactly what the studio produces.

Why

Most customers author in the studio, where an AI assistant does a large part of the flow-building for them and iteration is faster than any API call sequence could be. Programmatic authoring is what a minority of customers need, not the common path — so it isn't where the runtime API has been built out first.

If you need programmatic authoring

There's no client library for this either — no TypeScript or Python package. Every example on this site talks to /v1 with plain fetch or curl, and that's deliberate: it's what actually exists to call.

If your use case genuinely needs to create or edit agents from code rather than the studio, that's a different question from what /v1 answers today. Join the waitlist to talk about it.