Vapi
Stream meeting audio to your Vapi assistant and play its responses back into the meeting.
Keep managing your assistant, tools, and server webhooks in Vapi. Recall.ai creates a Vapi call using its WebSocket audio transport, sends the meeting audio mix, and plays the assistant's replies into the meeting.
Setup
1. Create or select a Vapi assistant
In the Vapi dashboard, open Assistants and select an existing assistant or choose Create Assistant. Configure its prompt, model, voice, and any server tools, then save it and copy its assistant ID. See the Vapi assistants quickstart.
Open API Keys → Private API Keys and create or copy a private key from the organization that owns the assistant.
2. Add your credentials to Recall.ai
Open Agents → Vapi in the Recall region where you create bots:
Select the workspace used by your Recall API key, then save your Vapi private key. If your assistant requires authentication for live call control, also save its organization's public API key in the optional field. Recall.ai uses live call control to end the Vapi call when output stops.
Credential changes apply to new connections.
Quickstart
Call Create Bot with your meeting URL and assistant ID. Replace the example UUID with your Vapi assistant's ID.
We recommend setting the relevant meeting platform in variant to web_4_core when pairing this audio agent with a video provider or a kind: "webpage" source. See bot sizing guidance.
{
"meeting_url": "MEETING_URL",
"output_media": {
"audio": {
"kind": "vapi",
"config": {
"assistant_id": "11111111-1111-4111-8111-111111111111"
}
}
}
}Once the bot joins the meeting, Recall.ai starts a Vapi call with your assistant. Your Vapi server URL and tool configuration continue to handle the assistant's business logic.
For recording the agent's voice, see Recording meeting agents. For response controls, see the Meeting agents FAQ.
Additional configurations
The output_media.audio.config object supports:
| Field | Required | Description |
|---|---|---|
assistant_id | For a saved assistant | UUID of an assistant in your Vapi organization. See the assistant setup guide. Mutually exclusive with assistant. |
assistant | For an inline assistant | An object matching the assistant schema in Vapi's Create Call reference. Mutually exclusive with assistant_id. |
assistant_overrides | No | An object matching assistantOverrides in Vapi's Create Call reference, such as a first message or variable values. See the assistant overrides guide. Recall.ai adds the reserved bot variables described below. |
Supply exactly one of assistant_id or assistant. Authentication comes from the dashboard's saved credentials; api_key, public_api_key, and credential_id are not accepted in this config.
Customize a saved assistant for a meeting
For example, use this config to set a first message and supply a value for a {{meeting_topic}} variable used by your assistant:
{
"assistant_id": "11111111-1111-4111-8111-111111111111",
"assistant_overrides": {
"firstMessage": "Hi, I'm ready to help with product planning.",
"variableValues": {
"meeting_topic": "Product planning"
}
}
}Recall.ai maps assistant_id to Vapi's assistantId and assistant_overrides to assistantOverrides. Fields inside the provider objects retain Vapi's names, including firstMessage and variableValues. See Vapi assistant overrides.
Use an inline assistant
To define an assistant for one call, supply assistant in place of assistant_id:
{
"assistant": {
"name": "Meeting assistant",
"firstMessage": "Hi, how can I help with the meeting?",
"model": {
"provider": "openai",
"model": "gpt-4o-mini",
"messages": [
{
"role": "system",
"content": "You are a helpful meeting assistant. Keep spoken replies brief."
}
]
}
}
}You can include voice, transcriber, and server settings supported by Vapi. See Transient vs permanent configurations for the assistant object.
Recall.ai manages the audio transport, enables monitorPlan.controlEnabled, and adds the speech-update, status-update, and user-interrupted client messages needed for playback and lifecycle handling. Server URLs and tool settings are preserved.
Use Recall bot context in tools
Recall.ai automatically adds these variables to assistantOverrides.variableValues in the Vapi call request, for both saved and inline assistants:
| Variable | Value |
|---|---|
recall_bot_id | The Recall bot's ID, matching the ID returned by Create Bot. |
recall_api_base_url | The bot's regional Recall API base URL, without an API version path, such as https://us-west-2.recall.ai. |
They are available through Vapi's variable templating, for example {{recall_bot_id}} and {{recall_api_base_url}}. Use them to associate your server tools with the meeting bot. Your tool server still needs your Recall API key to authenticate Recall API requests; the variables contain no Recall credentials.
Recall.ai supplies these variables even when you omit assistant_overrides or variableValues. Their names are reserved: Recall.ai replaces any values you supply for them while preserving your other variables and overrides. You do not need to add them to the bot request.
Append the endpoint's versioned path to this base URL when making Recall API requests.
Add a camera or avatar
Vapi is supported under output_media.audio. You can combine it with an independent camera or screenshare source. When audio is configured, the meeting audio mix goes only to the audio assistant; camera and screenshare sources do not receive it.
To use an automatic avatar, combine output_media.audio with automatic_video_output, leaving output_media.camera and output_media.screenshare absent.
FAQ
Do I need a phone number or Daily room?
No. Recall.ai connects through Vapi's WebSocket audio transport. The meeting URL belongs in the Recall.ai bot request.
Where should I implement tool calls?
Configure tools and server webhooks in Vapi or in the inline assistant. Vapi communicates with your server directly. Recall.ai does not execute browser-side tool handlers.
To add chat messages, screensharing, and a leave action, see Add meeting tools.
Updated about 2 hours ago
