Slack AI agents: a bot user your whole workspace can DM or mention

What you can build

How to connect Slack

1. 01On the Integrations page, copy the AgentRow manifest and create your Slack app from it: bot scopes, the app_mention and message.im events and the receiver URL are already filled in. Install it to the workspace.
2. 02Paste the Bot User OAuth Token and the Signing Secret under Integrations → Slack workspaces. AgentRow checks the token against Slack and keeps the workspace on the integration; credentials never leave the API except per run.
3. 03On the app's Workflow tab, put Slack (in) before Chat (in) and Slack (out) after Chat (out), pick the workspace on both (or create it right from the node), choose the triggers (DMs, mentions) and publish.
4. 04Give Slack the receiver URL from the app's Credentials tab if you did not use the manifest. Slack's URL check is answered automatically.
5. 05If the agent's tools act with an end-user access token, paste a personal access token as the workspace's tool access token (Settings → Access tokens) so those tools bind on Slack turns.

Example tool definition

The shape a tool takes in the AgentRow builder API. Credentials are stored on the tool's auth settings, not in its configuration, so they never reach a prompt. Placeholders in double braces are filled by the model from the input schema.

[
  {
    "id": "slack_in",
    "type": "slack_in",
    "config": { "integration_id": 8, "triggers": ["direct_message", "app_mention"] }
  },
  {
    "id": "slack_out",
    "type": "slack_out",
    "config": { "integration_id": 8, "reply_in_thread": true }
  }
]

What the run trace shows

Every turn shows its channel on the Chat (in) step (source, thread, caller), and every delivery is its own step after Chat (out) with the request, the channel's response, timing and a named error when it failed. Inbound events are listed on the app's Credentials tab with their outcome: accepted, duplicate or rejected, each with its reason.