Govern AI agents on Slack
Route approval requests to Slack channels, receive real-time agent alerts, and manage governance workflows without leaving your messaging platform. The Slack pack delivers approval notifications with one-click resolve, agent health summaries, and policy violation alerts.
What this pack does
- Approval request notifications with one-click approve/deny
- Agent health and heartbeat alerts
- Policy violation notifications with context
- Channel-based routing by risk tier
Use cases
Route high-risk approval requests to #ai-approvals channel
Alert on-call when an agent fails a safety check
Notify compliance team of policy overrides
Quick setup
- 1Install the Slack pack: cordumctl pack install slack
- 2Configure your Slack webhook URL in pack config
- 3Map approval tiers to Slack channels
- 4Enable the pack in your Cordum dashboard
Frequently asked questions
Can I limit which Slack channels agents are allowed to post to?
Yes. The Slack worker enforces a channel allowlist, so agents can only post to the channels you approve. Read actions are permitted by default, while write actions like posting messages run through policy and can be scoped per channel and risk tier.
How do approval requests reach the right people in Slack?
When the Safety Kernel returns a Require Approval decision, Cordum routes a notification to the channel you mapped to that approval tier, complete with the action context. Reviewers approve or deny inline, and the decision flows back to the gated job before it dispatches.
Will the Slack pack notify us when an agent fails a safety check?
Yes. Policy violations and failed safety checks are delivered as alerts with the originating agent, action, and policy reason. You can route these by risk tier so high-severity violations land in your on-call channel rather than a general feed.
What gets recorded when an agent posts to Slack?
Every Slack action and its governance decision (Allow, Deny, Require Approval, or Allow with Constraints) is written to the Cordum audit trail. Each result is audit-friendly, so you can trace which agent posted what, to which channel, and whether a human approved it.
Ready to govern Slack?
Other integrations
Govern AI agent actions on GitHub repositories.
Govern AI agent read and write actions against the AWS API.
Governance for AI agents managing Jira workflows.
Govern AI agents responding to Kubernetes incidents.
Feed Datadog alerts into governed agent workflows.
Govern AI agents in PagerDuty incident workflows.