Send Recall.ai meeting bots to Zoom, Google Meet, Teams and Webex calls, and read back recordings, transcripts, participant events and calendar events.
Recall.ai ships in the w6w first-party pack. It declares 29 actions, 3 health checks, and the host runs its code in a sandbox that never sees the credential.
io.w6w.recallaiRecall.ai is an API for meeting bots: a bot joins a Zoom, Google Meet, Microsoft Teams, Webex or GoTo meeting, records it, and hands back the video, audio, transcript and participant data. This app connects a workflow to the Recall.ai REST API with an API key, in any of Recall’s four regions (US East, US West, EU, Asia).
A workflow can create a bot now or at a scheduled time, update or delete a scheduled bot, remove a bot from a call, make it post to the meeting chat or play audio, and start, pause, resume or stop its recording. It can list and fetch bots, recordings, transcripts, participant events and meeting metadata, run an asynchronous transcript on a finished recording, and delete recordings and transcripts. For Calendar V2 it can list calendars and events and schedule or unschedule a bot for an event. It can also read the workspace’s bot usage.
Recall hosts each region separately, so the connection records the region and every call goes to that host. Errors are reported with Recall’s own code. The OAuth secrets Recall echoes back on calendar objects are removed. Health checks verify the key and whether the API answers.
Three routes to the same 29 actions. The Workflow tab is generated from Recall.ai's own manifest and carries its real ids, so it is copy-pasteable; the Code and CLI examples are the same call for any action on any app, so every app-specific value in them is a blank you fill in.
bot-create Send a bot to a Zoom, Google Meet, Microsoft Teams, Webex or GoTo meeting to record it, now or at a scheduled time. Transcripts, recordings and participant events are read back after the call.
bot-delete Delete a bot that has not joined a call yet. A bot already dispatched answers 405; use Remove Bot From Call instead.
bot-delete-media Delete the media Recall stored for this bot (recordings, transcripts, participant events). Irreversible. Returns the bot.
bot-get Retrieve a bot: its status history, recordings and the media shortcuts (transcript, video, audio) with their download URLs once ready.
bot-leave-call Remove the bot from the meeting. Irreversible: the bot leaves and processes what it recorded. Fails with `cannot_command_completed_bot` or `cannot_command_unstarted_bot` when the bot is not in a call.
bot-list List bots in the workspace, filtered by join window, meeting URL, platform, status or a metadata key. Not for polling a bot's status: use Recall's status-change webhooks for that.
bot-output-audio Make the bot play an MP3 into the meeting. The bot must have been created with `automatic_audio_output` configured (set it in Other bot settings on Create Bot).
bot-send-chat-message Make the bot post a message in the meeting chat. The bot must be in the call.
bot-start-recording Instruct a bot already in the call to start recording, optionally with its own recording config. Restarts the current recording if one is in progress.
bot-stop-recording Instruct the bot to end the current recording while it stays in the call.
bot-update Change a scheduled bot's meeting URL, name, join time, recording config or metadata before it joins the call.
calendar-event-get Retrieve a calendar event, including the bots scheduled for it.
calendar-event-list List the events of one calendar, with the bots scheduled for each.
calendar-event-schedule-bot Schedule a bot to join a calendar event when it starts, or update the bot already scheduled for it. Events sharing a deduplication key share one bot.
calendar-event-unschedule-bot Remove the bot scheduled for a calendar event so it does not join.
calendar-get Retrieve a connected calendar (Calendar V2). The OAuth refresh token and client secret Recall echoes back are removed.
calendar-list List connected calendars (Calendar V2). The OAuth refresh token and client secret Recall echoes back are removed from every item.
meeting-metadata-list List meeting-metadata artifacts (for Zoom, the meeting UUID).
participant-events-list List participant-event artifacts. Each carries `data` download URLs for the raw events, the speaker timeline and the participants list.
recording-create-transcript Transcribe a finished recording with an asynchronous provider. Returns a transcript in `processing`; fetch it again until its status is `done`.
recording-get Retrieve a recording and its media shortcuts (mixed video and audio, transcript, participant events, meeting metadata).
recording-list List recordings, newest first, filtered by bot, creation window or status. Each carries media shortcuts.
transcript-get Retrieve a transcript. When its status code is `done`, `data.download_url` is a pre-signed link to the transcript JSON.
transcript-list List transcripts, filtered by recording, creation window or status.
A workflow step names the app and the action, and the editor fills in the
connection when you pick one. This is the Step shape from the
workflow spec, carrying Recall.ai's real ids.
{
"manifestVersion": "2",
"name": "recallai-example",
"steps": [
{
"id": "bot-create",
"uses": {
"app": "io.w6w.recallai",
"action": "bot-create",
"connection": "conn_YOUR_CONNECTION_ID"
},
"with": {
"meetingUrl": "<meetingUrl>"
}
}
]
}bot-create bot-get bot-list bot-output-audio bot-pause-recording +24 more actions available
Every app-specific value here is a blank you have to fill in. An
app action is reached through the connection that authenticates it, so the
address is a connection id, not the app id — and connections belong to your account,
so a public page cannot know yours. Create one for Recall.ai, then fill in
the three blanks: conn_YOUR_CONNECTION_ID, the action key, and the
parameters that action declares. The call itself is real — the shape is transcribed
from the studio's own snippet builder, which prints the same kind of blanks — but
nothing in it is specific to Recall.ai. The Workflow tab is where this app's
real ids are.
npm install @w6w/sdkyarn add @w6w/sdkpnpm add @w6w/sdkdeno add npm:@w6w/sdkimport { W6wClient, isActionRun } from "@w6w/sdk";
// Reads W6W_BASE_URL and W6W_TOKEN from the environment when omitted.
const client = new W6wClient();
const envelope = await client.run({
urn: "conn_YOUR_CONNECTION_ID",
action: "bot-create",
payload: {
meetingUrl: "<value>",
// botName: "<value>",
// joinAt: "<value>",
// transcriptProvider: "<value>",
// transcriptLanguage: "<value>",
// recordingConfig: "<value>",
// extra: "<value>",
// metadata: "<value>",
},
});
if (isActionRun(envelope)) console.log(envelope.value); npm install -g @w6w/cli w6w run conn_YOUR_CONNECTION_ID --action bot-create --payload '{"meetingUrl":"<value>"}' Give an AI agent Recall.ai — without giving it Recall.ai's credentials. One MCP endpoint exposes every app, function and workflow the caller is entitled to, as tools it can discover and run. Access is granted per team while we onboard.
One tool call{
"name": "w6w_invoke",
"arguments": {
"ref": "app:io.w6w.recallai#bot-create",
"input": {
"meetingUrl": "<meetingUrl>"
}
}
}
Every tool names its target with a single ref. The
app: form above doesn't name a connection at all — the
host resolves which of the caller's Recall.ai connections to sign
with, and refuses rather than guesses when the answer is ambiguous.
The token is attached host-side, at the moment of the call. It is never a tool argument, never in the model's context, and never in a transcript — so a prompt injection has nothing to exfiltrate.
Tools are derived per end user from what that person has actually connected and is entitled to — not one shared bot identity carrying the union of everyone's access.
Multi-step work runs on the workflow engine and returns a run handle the agent can poll — retries, branching and state survive the conversation that started them.
Recall.ai's declared health checks are on the surface too, so an agent can tell "the vendor is down" from "your credential expired" before it burns a retry on either.
The MCP surface is part of the hosted platform. Recall.ai itself is MIT, and the runtime that executes it is source-available (FSL).
Recall.ai declares its own checks, so its health is a property of the app rather than something the host guesses at.
Unauthenticated GET of /api/v1/bot/ on the connection's region host. Recall's documented 401 `not_authenticated` error is the expected healthy answer; key validity is the `auth:api-key` check's job.