Skip to content

AI Agents

AI agents are configurable assistants that live inside Bruin Cloud. Each agent can be scoped to a project, attached to a connection set, connected to messaging platforms, and given a custom system prompt. The same agent can be used in chat, embedded in Slack, Microsoft Teams, Discord, WhatsApp, or Telegram, scheduled to run on a cadence, or used to build dashboards.

How agents work

When you send an agent a prompt, it:

  • Spins up a sandbox environment.
  • Clones the repo of the project it is connected to (if any).
  • Reads your pipelines, assets, and any AGENTS.md or instruction files.
  • Builds a memory of that context so it can answer questions in the right scope.
  • Uses the agent's connection set to query the data warehouse, typically starting by inspecting the schema to map out available tables.

If the agent has Cloud CLI access enabled, it can also operate Bruin Cloud itself: read pipeline run history, inspect assets, trigger pipelines, and query the catalog and glossary.

For how the underlying data is scoped and retained, see Does the agent see my actual data? in the FAQ.

What is in this section

  • Configure Agents: create an agent, pick a project, attach a connection set, add messaging integrations, set permissions.
  • Chat with Agents: use the agent in the Bruin Cloud chat for analysis, reporting, and CLI tasks.
  • Scheduled Agents: run an agent on a cadence (daily reports, threshold alerts, custom SQL runs).
  • Integrations: connect an agent to Slack, Microsoft Teams, Discord, WhatsApp, or Telegram so your team can query data from where they already chat.
  • Slack AI Analyst tutorial: end-to-end walkthrough that builds a pipeline, enhances metadata, and deploys an analyst to Slack.

Where agents fit

Use caseWhere
Ask one-off data questionsAI → Chats
Embed answers in your team's chat toolMessaging integrations (Slack, Teams, Discord, WhatsApp, Telegram)
Generate dashboards from promptsAI → Dashboards
Send daily/weekly reports automaticallyAI → Scheduled Agents
Manage pipelines from the terminal via AICloud MCP

When an agent reads your repo, it interprets it through the same primitives the CLI uses: