---
title: Slack
summary: Connect a Slack workspace so your AI agents can read channels, search messages, and post and reply under permissions you control.
category: integrations
slug: slack
---

## What Slack in Kronos lets your agents do

Once a Slack workspace is connected to Kronos, your agents can — within
the permissions you grant — work with the channels, conversations, and
people in that workspace:

**Read actions** (available when an agent has "Read only" or higher):

- List the channels in the workspace and read a channel's details.
- Read the recent messages and full history of a channel or thread.
- Search messages across the workspace.
- Look up the people in the workspace and read a user's profile.
- Read the connected workspace's own profile and team information.

**Write actions** (available only when an agent has "Read and write"):

- Post a message to a channel or send a direct message.
- Reply inside a thread.
- Update or delete a message the connection posted.
- Add an emoji reaction to a message.
- Create a channel, and invite or remove people from one.

You decide which agent gets which level, and you can switch any
permission off in one click without disconnecting the Slack workspace.

## Connecting Slack

1. Open Kronos → **Connect** in the sidebar → **Applications** tab.
2. Click the **Slack** card to open the detail dialog, then click the
   green **Connect** button.
3. You land on Slack's authorization screen. Sign in if you aren't
   already, and pick the workspace you want to connect.
4. Review the access Slack lists and approve it.
5. You're sent back to Kronos at the **Already connected** tab with the
   new Slack entry pre-selected.

You can connect more than one Slack workspace; each appears as its own
connection with its own workspace name.

## Telling connections apart

The Slack management page shows the connected workspace's name right
under the "Slack" title (for example, `Acme Inc`). This is how Kronos
and you identify which workspace this connection represents.

If you connect more than one workspace, each appears with its own
workspace name. Optionally you can add a short personal note next to it
(the "Your note for this connection" field) — labels like "Company HQ",
"Support team", or "Client space" make it easier to scan a long list.
The note is yours only; agents always identify the workspace by its name
and address, not by the note.

## Giving an agent access

On the Slack management page click **+ Add AI Agent**, pick the agent
from the dropdown, then choose:

- **Off** — the agent cannot touch this Slack workspace.
- **Read only** — the agent can list channels, read messages, and search
  conversations, but cannot post, reply, or react.
- **Read and write** — the agent can do all of the above plus post
  messages, reply in threads, react, and manage channels.

The same controls are available inside the AI Agent editor — scroll to
the **Connected Apps** section.

## Fine-grained overrides

Inside the permissions modal (the gear icon next to a Slack workspace in
the agent editor or on the management page) you'll see the full list of
Slack actions grouped by **Read tools** and **Write tools**. Toggle
individual actions on or off to override the preset.

Common combinations:

- **Monitoring-only agent**: preset Read only, no overrides. The agent
  reads channels and summarizes activity but cannot post anything.
- **Alerting agent**: preset Read only, override `Send a message to a
  channel` ON. The agent reads conversations and can post alerts, but
  cannot delete messages or manage channels.
- **Safe responder agent**: preset Read and write, override `Delete a
  message` OFF and `Create a channel` OFF. The agent can post and reply,
  but cannot remove messages or restructure the workspace.
- **Full assistant**: preset Read and write, no overrides.

Click Save to apply. Changes take effect on the next message the agent
processes.

## Multiple Slack workspaces on the same agent

If an agent has access to two or more Slack workspaces (e.g. Company HQ +
Client space), the agent picks which workspace to use on every action
based on context. When the request is ambiguous ("post this in
#general"), the agent will ask which workspace to use.

You can cut ambiguity by turning the agent's access to one workspace
off, or by giving the connections obviously distinct notes ("Company HQ"
vs "Client space").

## When something goes wrong

- **"Needs reconnect"** badge — Slack revoked or expired the connection.
  Open `/app/connect`, click Slack, walk through Connect again.
  Permissions are preserved.
- **Agent says "I can't find that channel"** — the connection isn't a
  member of that channel. In Slack, invite the connection to the channel
  (private channels require an explicit invite), then try again.
- **Agent says "I can't post to that channel"** — that agent's
  permission preset is Off or Read only without the matching write
  override. Open the permissions modal and adjust.
- **Agent says the connected app is temporarily unavailable** — Slack or
  the connection had a transient hiccup. Ask the user to retry in a
  moment. If it persists, check the management page for a "Needs
  reconnect" badge.

## Privacy

Kronos does NOT store the content of your Slack messages. Each time an
agent uses a Slack action, it fetches just-in-time from Slack through the
connection, uses what it needs to complete the task, and discards the
result. The audit log records that the action happened (which agent,
which action, when, and whether it succeeded), but not the message
content.

When you disconnect Slack or cancel your Kronos subscription, the
connection and all per-agent permissions are deleted immediately.
