Channels you're in.
Public and private channels your Slack account is a member of. Threads, replies, and the surrounding message context.
Slack integration
Chat with Slack.
The decision happened in a thread three weeks ago. Chat with Work searches Slack the way you would — find the right messages, open the surrounding thread, and answer with the conversation linked back.
Read-only, live today
It searches like a person would.
No background indexing of your workspace. No firehose of every message we can see. Chat with Work pokes at Slack the same way you would: find the hit, open the thread around it, then read.
-
Step 1
You ask a question.
Plain English. No channel filter, no quoted phrase rules.
-
Step 2
It searches Slack.
Behind the scenes, it queries Slack's search API and gets a small set of hits back — channel, author, time, and the message text around your keywords.
-
Step 3
It opens the right threads.
For the hits that look promising, it pulls a few messages before and after and the thread replies, so the answer reflects the full exchange — not just the matching line.
-
Step 4
You get an answer with the source.
The reply names the channel, the people, and the time, and links straight to the original message in Slack.
Searches stay scoped to whatever your connected Slack account can already see. If you can't read it in Slack, neither can Chat with Work.
What it can actually see.
The same things you can. Nothing more.
DMs and group DMs.
Direct messages and multi-person DMs your account already has access to. Useful for customer threads and one-on-ones.
People and channel names.
User and channel metadata so results read like "Sara in
#launch" instead of U02ABC123.
Nothing gets posted. The integration only reads. There's no write permission in the OAuth request.
What people actually ask
The Slack questions that usually turn into channel digging.
What did we decide about the launch?
Who is waiting on me?
Find the thread about the invoice issue.
What changed since the last customer call?
Which channel did the SOC 2 discussion happen in?
Who flagged the security incident first?
Read-only, scoped to your view.
The integration asks only for read access on conversations your user can already see. No write permission, no posting, no bot spamming your channels.
Messages that the model pulls to answer a question are encrypted at rest, scoped to your user, and deleted after 30 days if nothing touches them again. Nothing in your account is used to train models.
Full security write-up on the security page.
Disconnect any time.
- From Chat with Work: Settings → Connected accounts → Disconnect. Tokens are deactivated and cached messages queued for deletion.
- From Slack: workspace admins can revoke under Apps → Manage. We detect it on the next call and stop.
Ready to connect Slack?
Start free with Google, then add Slack from settings when you want conversations in the same assistant.
Create your account, then add Slack from settings.