Platform · Slack

Observe AI agents that work in Slack.

Your agent reads channels, sends messages, replies in threads, uploads files, reacts to posts. Trefur captures each Slack call as a span on the agent trace — without changing the integration code you already wrote.

What you see.

Channel reads: which channel, how many messages returned, filters applied.
Message sends: channel or DM, recipient, length, timestamp, success or failure.
Thread replies: parent thread ID, reply count, attachments attached.
File uploads: file type, size, channels shared into.
Reactions: emoji, target message, time added.
Slash command handling: command name, args, user who triggered it.
Webhook deliveries: target URL, status code, retry attempts.

Why it matters.

For engineers

Each Slack Web API call your agent makes becomes a tool span. When an agent posts to the wrong channel or pings the wrong user, you see it on the trace before the screenshot lands.

For platform teams

Roll up Slack calls per agent, per workspace, per environment. Spot agents posting outside their declared channels. Audit which bot identities are most active.

For business operators

Know what your AI is saying in Slack and to whom. Tie response volume to ticket resolution. Catch silent stalls where agents stop replying mid-thread.

For SRE

Slack rate-limit errors and webhook failures land on the span. Envelope alerts catch retry-rate spikes per workspace. Oncall opens the trace pre-triaged.

How it works.

Your agent calls the Slack Web API to do its job. The Trefur SDK wraps the client and emits a tool span per call — chat.postMessage, conversations.history, reactions.add, and the rest. Spans land on the agent trace alongside the LLM call that decided to post, so the full sequence is one navigable timeline. The collector can run inside your network if you want message bodies to stay local.

See what your Slack agents are actually doing.

Free tier. No card required. First trace in under five minutes.