Read Fellow meeting notes, recordings, transcripts and action items, edit note agendas, upload recordings and manage webhooks.
Fellow ships in the w6w first-party pack. It declares 23 actions, 2 health checks, and the host runs its code in a sandbox that never sees the credential.
io.w6w.fellowFellow is an AI meeting assistant: it joins and records meetings, transcribes them, writes the notes and pulls out the decisions and action items. This app puts that output into your workflows through Fellow’s Developer API.
Read notes and recordings (with transcripts and AI-generated sections on request), list and filter them by meeting, channel, attendee or date, and walk the results page by page. Action items can be listed, completed, reopened and archived, so a workflow can push a meeting’s follow-ups into a task tracker and close them back out. Note agendas can be replaced, appended to, prepended to, or edited with an exact find-and-replace, and notes renamed.
Recordings made elsewhere can be sent to Fellow for transcription from a media URL, with the import status tracked afterwards. Webhooks for new AI notes, notes shared to a channel, and action items assigned or completed can be registered and managed, which is how a workflow reacts to a meeting ending instead of polling for it.
A Super Admin key additionally unlocks workspace-wide reads, acting on behalf of another user, workspace-scoped webhooks and permanent deletion of notes and recordings. Each Fellow workspace has its own host, so the connection takes your workspace subdomain.
Three routes to the same 23 actions. The Workflow tab is generated from Fellow'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.
action-item-list List action items with optional filters and ordering, one page at a time.
me-get Get the user behind the API key and the workspace it belongs to. A cheap way to confirm a connection and to read the workspace's id and subdomain.
note-agenda-append Add Fellow Markdown to the end of a note's agenda, leaving the rest in place.
note-agenda-find-replace Replace the single occurrence of a string in a note's agenda with Fellow Markdown. The text must match exactly once.
note-agenda-prepend Add Fellow Markdown to the start of a note's agenda, leaving the rest in place.
note-agenda-write Replace a note's whole agenda with Fellow Markdown, and/or rename the note. Give content, a title, or both.
note-delete Permanently delete a note AND all of its recordings. Requires a Super Admin key.
note-list List notes with optional filters, one page at a time. Note bodies and attendees are opt-in.
recording-delete Permanently delete a recording. Requires a Super Admin key; deleting a recording does not delete its note.
recording-list List recordings with optional filters, one page at a time. Transcripts and AI notes are opt-in because they are large.
recording-upload Queue a recording for import and transcription from a media URL Fellow can fetch. Returns the new recording id; poll Get Upload Status for the outcome.
upload-get Retrieve one API-uploaded recording with its import status (PENDING, SUCCESS or FAILED) and failure reason.
webhook-create Register a webhook. Fellow immediately POSTs a `url_verification` challenge to the URL, and the webhook is only created if the endpoint echoes the challenge back as plain text. The signing secret is returned once, here.
webhook-list List webhooks, one page at a time. A non-privileged key lists its own; a Super Admin key lists the whole account.
webhook-update Partially update a webhook: only the fields you give are changed. Changing the URL triggers a fresh url_verification challenge. Scope cannot be changed.
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 Fellow's real ids.
{
"manifestVersion": "2",
"name": "fellow-example",
"steps": [
{
"id": "action-item-complete",
"uses": {
"app": "io.w6w.fellow",
"action": "action-item-complete",
"connection": "conn_YOUR_CONNECTION_ID"
},
"with": {
"actionItemId": "<actionItemId>"
}
}
]
}action-item-complete action-item-get action-item-list me-get note-agenda-append +18 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 Fellow, 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 Fellow. 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: "action-item-complete",
payload: {
actionItemId: "<value>",
// completed: "<value>",
// onBehalfOf: "<value>",
},
});
if (isActionRun(envelope)) console.log(envelope.value); npm install -g @w6w/cli w6w run conn_YOUR_CONNECTION_ID --action action-item-complete --payload '{"actionItemId":"<value>"}' Give an AI agent Fellow — without giving it Fellow'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.fellow#action-item-complete",
"input": {
"actionItemId": "<actionItemId>"
}
}
}
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 Fellow 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.
Fellow'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. Fellow itself is MIT, and the runtime that executes it is source-available (FSL).
Fellow declares its own checks, so its health is a property of the app rather than something the host guesses at.
The Developer API component on status.fellow.ai, with the other Fellow components (application, website, MCP, bot recording, summarization) reported as detail.