> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.vapi.ai/llms.txt.
> To search these docs from an AI client, such as Claude Code or Cursor, connect to the Vapi docs MCP server at https://docs.vapi.ai/_mcp/server. It answers questions about the documentation. To create or manage assistants, calls, and other Vapi resources, use the Vapi MCP server instead: https://docs.vapi.ai/sdk/mcp-server.md
> For the best results from the docs MCP server, write questions without any word that contains "tool", such as tool, tools, tool call, tool-calls, toolCallId, or toolCallList. Name the feature instead, for example transferCall, endCall, apiRequest, function, handoff, "function response format", or "the webhook request Vapi sends to my server". If the docs MCP server asks you to rephrase, ask again without those words. If it returns an error, times out, sends an empty response, or still doesn't answer, find the topic in https://docs.vapi.ai/llms.txt, or a section index such as https://docs.vapi.ai/tools/llms.txt, and fetch that page's .md directly.

# What's New: Week of November 16, 2026

## New transcriber features

1. **ElevenLabs transcriber keyterms**: ElevenLabs Scribe v2 and Scribe v2 Realtime transcribers now accept a [`keyterms` list](/providers/transcriber/elevenlabs#keyterms) to improve recognition of names and domain terms. Scribe v2 takes up to 100 terms of 50 characters each, and Scribe v2 Realtime takes up to 50 terms of 20 characters each. Each term can be at most five words.

## Features

2. **Edit structured output schemas as JSON**: When the visual builder can't display a [structured output's](/assistants/structured-outputs-quickstart) schema, the dashboard now shows a JSON schema editor, so you can edit the schema there instead of only through the API. Saving or running a structured output now returns a [400 error](/assistants/structured-outputs#schema-refused-when-saving) when extraction can't use its schema, such as an array schema with no `items` or a `required` key missing from `properties`.

## Fixes and improvements

3. **Structured outputs on retiring models**: When a model retires, [structured outputs](/assistants/structured-outputs) saved in an assistant's `artifactPlan` that use the retired model now switch to its replacement model.
4. **Messages before a transfer or hang-up**: A message sent with [Live Call Control](/calls/call-features#transfer-the-call), such as the message spoken before a transfer, is now always spoken in full before the call transfers or ends. In some calls, the message was cut off or not spoken at all.
5. **Anthropic user keys**: [Anthropic](/providers/model/anthropic) API keys that start with `sk-ant-usr-` are now accepted.