I tried using SuperAGI to spin up a Slack bot last week. The goal was simple: automate a daily digest from our CRM and post it to a channel.
The core Slack integration via the toolkit is straightforward. You plug in the Slack Bot Token and Signing Secret in the config. The real challenge is getting SuperAGI to handle the conversational logic reliably. Out of the box, it wants to treat every Slack message as a new task, which can get messy fast.
What ended up working was a very specific agent workflow. I created an agent dedicated *only* to listening for a specific slash command (like `/digest`). That agent triggers a separate, internal tool that fetches the CRM data, formats it, and uses the Slack toolkit to post. Trying to make one agent handle open-ended Slack conversations and the data-fetching kept failing. The action space got too noisy.
So my takeaway: use SuperAGI to manage the workflow and call your tools, but keep the Slack-facing agent's role extremely narrow. Offload the complex logic to custom tools outside the agent's main loop. Has anyone else tried a similar approach? Curious about handling mentions or DMs without it going off the rails.