n8n lets you describe an agent once and reuse it in Slack or on a schedule

On 25 September 2026, n8n launched Agents as a first-class object, not only a node on the workflow canvas. You describe what the agent should do, pick a model, and choose which tools and workflows it may use. The agent then works out the steps. The same agent can be talked to in Slack, run on a schedule, or called from any workflow. n8n says this is meant for open-ended or back-and-forth jobs that are hard to lock into a fixed path. The older AI Agent node remains. Agents are in preview on n8n Cloud for users on the latest stable version. Self-hosted setups need extra work, and Enterprise is still coming. One conversation turn counts as one execution. Calls into workflows and sub-agents do not add extra counts, and agents share the workflow execution quota. Building with n8n Assistant uses AI credits. n8n told users to test before publishing and keep approvals on anything sensitive.
Until now, a reusable AI job in n8n usually meant drawing a fixed path on the canvas, or dropping an AI Agent node inside that path. Chat, schedule, and workflow were separate designs. If a task needed back-and-forth judgement, you either overbuilt the map or kept the work in a one-off chat that nobody else could run. The change is that the agent is now the reusable unit. You write the job once, limit it to workflows you already trust, and meet it in Slack or on a timer without redrawing the path. That only holds if the description, allowed tools, and human review stay explicit. Preview status and extra self-hosted setup mean this is not yet a default team system.
Analysis
Treat this as a test, not a rollout. Pick one messy, back-and-forth task you already do by hand, write a short job description, attach only existing workflows as tools, leave approvals on, and try the same agent in Slack before you put it on a schedule.
Source note
Pulse published by Collab365 Spaces, reviewed by Helen Jones on . Cite as "n8n lets you describe an agent once and reuse it in Slack or on a schedule", Collab365 Spaces. 2 sources referenced.