Parker works from the conversations you already have in Slack — so the first thing a firm does is install it as a bot and grant it a small set of scopes through Slack's developer portal. You'll register an app, add permissions, and generate one token to hand off securely. Nothing about your existing Slack setup changes; Parker simply gains the ability to read and respond in the channels you choose.
Before you begin. You'll need permission to install apps in your firm's Slack workspace — usually a workspace owner or admin. Open api.slack.com/apps in another tab and follow along — by the end you'll have one token to hand to the Parker team, and your firm will be ready to connect.
The seven steps
- 01 Go to the Slack API siteGo to api.slack.com/apps and sign in with your normal Slack account. You'll land on the Your Apps screen.
- 02 Create a new appClick Create New App, choose From scratch, name it Parker, and select your firm's workspace from the dropdown.
- 03 Add the Bot Token Scopes Parker needsUnder OAuth & Permissions → Scopes → Bot Token Scopes, add the scopes below — this is the only access Parker requests. See the breakdown further down.
-
04
Turn on Event SubscriptionsUnder Event Subscriptions, toggle it on and subscribe to
message.channelsandmessage.im, then save. This lets Parker notice when it's mentioned or messaged. - 05 Install the app to your workspaceUnder Install App, click Install to Workspace, review the permissions, and click Allow.
-
06
Copy the Bot User OAuth TokenOn the Install App page, copy the token beginning with
xoxb-, then submit it securely to the Parker team — that's how you hand off the connection. -
07
Invite Parker to a channelOnce we've confirmed the connection, open any channel and type
/invite @Parker. Add or remove it from channels at any time.
A closer look at permissions
Step 3 is the one worth slowing down on. Slack's permission model is scope-based — you grant a bot exactly the capabilities it needs, one at a time, rather than an all-or-nothing connection to the workspace. Parker doesn't ask for the ability to manage members, billing, or other apps, and it doesn't get a feed of every conversation in the workspace either.
The principle is simple: grant access to the channels Parker is actually invited into, and nothing about the workspace beyond that. Membership, billing, and workspace administration have nothing to do with answering a question about a matter — so they stay out of scope entirely.
Grant these bot scopes
These are the scopes Parker uses to read and respond in the channels it's added to:
- channels:history — read messages in public channels Parker has been invited to.
- groups:history — read messages in private channels Parker has been invited to.
- channels:read — see basic information about public channels.
- chat:write — send messages and replies as Parker.
- im:history — read and respond to direct messages sent to Parker.
- users:read — match names and @mentions to the right person.
- files:read — read files shared in channels Parker has been invited to.
Leave out entirely
These have nothing to do with drafting or answering from a matter, so Parker's app never requests them:
- Admin scopes — workspace settings, billing, and app management.
- Full workspace history — Parker reads only the specific channels it's invited to, never a firm-wide feed.
- Member management — adding, removing, or changing the role of anyone on the team.
Why restrict at all? Channel-scoped access already keeps Parker out of conversations it hasn't been invited into. Skipping admin and workspace-wide scopes goes a step further — it shrinks the blast radius. Even in the unlikely event a token leaked, there would be no path to workspace settings, billing, or channels Parker was never added to, because that access was never granted.
What happens next
Once you've handed off the token, the Parker team completes the connection on our side. From there, Parker works from your matters inside the tools your firm already uses — Slack, Telegram, Outlook, and more — across whatever practice areas you handle. There's no platform to migrate onto and nothing new for your team to learn.
And because every firm runs in an isolated environment with Zero Data Retention, your clients' information is never stored by the model provider or used to train any model. Attorney–client confidentiality was a design constraint from day one.