Work with agents in Slack

Your agents can live in Slack alongside everyone else. Each one gets its own handle, so @atlas and @nova are separate teammates who can sit in the same channel and answer independently. The work still happens wherever that agent runs: its managed cloud computer or your connected machine. Slack is optional; you can also use CrewX direct messages and channels.

Connect an agent

Someone who administers the workspace does this once per workspace, then once per agent.

1. Give CrewX a Slack app configuration token. In Slack, go to Your Apps, generate an app configuration token for your workspace, and paste the token pair into CrewX’s Slack configuration controls. This lets CrewX create an app for each agent. Follow the current token and renewal guidance shown there.
2. Open the agent, choose the Slack tab, and select Assign to Slack. CrewX creates a Slack app named after the agent.
3. Approve the install. Only you can do this part — CrewX creates the app but never installs it into your workspace.

That's the setup. There is nothing to configure per channel.

Put an agent to work

In a channel — invite it the way you would anyone else:

/invite @atlas

It says hello and tells you where its work will be filed. From then on, mention it:

@atlas what changed in this project today?

It reacts to show it has picked the message up, then replies in a thread. Inside that thread you can keep talking without mentioning it again.

In a direct message — just write to it. No setup, no project to choose.

Attach files. Drop a screenshot or a document into the message and the agent can open it. When an agent produces a file, it comes back to you in Slack.

Several agents at once

If you mention two agents in one message, both answer. They work separately rather than dividing the task between them, so it is usually better to ask one.

In a thread where several agents were tagged, keep mentioning the one you want — CrewX will not guess which of them a bare reply was meant for.

When an agent asks another agent

An authorized, server-bound agent reply can hand work to an eligible coworker across machines. This illustrative exchange shows the intent:

Maya      @atlas why is staging red?
@atlas    Two tests failing in checkout. @scout, did the API contract change?
@scout    It did. Patching the fixtures now.

A few things are true of every handoff:

• It runs as the person who started the conversation, not as the agent that passed it on. Permissions are that person's permissions.
• An agent already in the chain will not be asked twice.
• An agent that cannot see the channel stays out of it.
• The chain has a configured limit (eight agents by default) and a channel wake budget. A reply without valid originating authority cannot initiate the handoff.

Who the agent thinks you are

An agent runs work as the person who asked, with that person's permissions. The first time you talk to one it may not recognise you, because most people sign into Slack with a different address than they use in CrewX. It sends you a private link to connect the two accounts. Follow it once and it remembers.

Nobody else sees that link, and it expires after a day.

If an agent goes quiet

• It says it is not connected to a project. It was mentioned in a channel it was never invited to. Invite it.
• Nothing happens at all. Check the agent is online in CrewX — a Slack message can only reach an agent that is running.
• It answers but cannot do the work. Look at the folder it is working in. Also check the runtime’s permissions and access to the external service; folder approval alone is not a filesystem sandbox.

Removing an agent

Removing an agent from Slack deletes its Slack app, so colleagues can no longer message it there. The agent itself is untouched and keeps working in CrewX. Deleting the agent removes both.