Integration · Your day
Ask Brigham.ai where your team talks.
Slack is another place to reach Brigham.ai. The rules do not change with the surface.
What it does
01
Ask in a channel or DM
Questions get answers with sources, in the place the question came up.
02
Briefings where you read
Your morning briefing can arrive in Slack instead of a new tab.
03
Approve in place
An approval request can be answered from Slack. The approval engine still decides.
Scopes and access
- Connects as
- Each person, their own account
- Access
- Read-only by default
- Reads
- Channels and messages you grant
- Writes
- Posting messages (separate grant)
- Needs approval
- Posting to shared channels on your behalf
- Visibility
- Private channels stay private to their members
- Availability
- Available
Yours to connect, yours to revoke.
Each person connects their own accounts from the Connections page. Every connection can be revoked on its own, and every connection starts read-only. A write scope is a separate, explicit grant.
- Credentials are encrypted at rest and referenced by id. They never appear in model context, logs or tool results.
- Each agent gets its own connection grants.
- Tool results from outside sources are untrusted data. They cannot approve anything.
Ask it
- ›“What did the team decide about the launch date?”
- ›“Summarize #support from yesterday.”
- ›“Post my briefing to #leadership.”
- ›“Who asked me something I have not answered?”
Questions, answered.
Is Slack available now?
Yes. Connect it from the Connections page, with the same scopes and approvals as every other integration.
Can a Slack message approve an action?
No. Message text from Slack is untrusted data. Approval goes through the approval engine.
Keep reading.
GmailBrigham.ai searches and summarizes your Gmail live, finds the follow-ups you owe, and writes drafts into Gmail Drafts. Nothing is sent without approval.Read →
ApprovalsEvery approval is bound to the exact payload hash. Change the action and you approve again. Email links confirm on click, never on open.Read →
Untrusted contentEmail, web pages, documents and notes are labelled as data. Nothing that arrives through a tool can approve an action, change a policy or widen a scope.Read →