Skip to content

AI Assistant

BCP Portal ships with a built-in AI Assistant — a chat companion that answers cybersecurity questions in your own language, tailored to your knowledge level, using only the data you already have access to.

Powered by enterprise Claude

The Assistant is built on Claude by Anthropic, using the enterprise tier. Practical consequence:

  • Anthropic does not use your data to train its models.
  • Your conversations, alerts, and organization data don't leave the platform's isolated context.

Where you can talk to the Assistant

There are three entry points, all using the same underlying assistant:

  • Built-in chat inside the portal — the top bar has an AI Assistant button that opens a chat panel over any page.
  • Slack — through the platform's bot (/bcp_ask command). See Slack Integration.
  • MCP Server — connect the assistant to your Claude Desktop, Cursor, or other MCP client. See MCP Server.

Configuration

When you first open the chat you can choose:

  • Language — any language convenient for you (English, Ukrainian, or others). The Assistant switches to it and stays there for future sessions until you change it.
  • Cybersecurity knowledge level — beginner / intermediate / expert. Higher level → the Assistant uses more technical terms and skips introductory context. Lower level → the Assistant explains fundamentals along with recommendations.

These preferences can be changed anytime through the settings icon in the chat.

What the Assistant can do

  • Explain a specific alert: what does it mean, why is it Critical, what to do next.
  • Summarize a period: "what happened in security this week."
  • Recommend actions: "which of my active High alerts should I resolve first."
  • Compare states: "did my posture score improve compared to last month."
  • Guide setup: "how do I connect GitHub", "what should I put in a scheduled report."

Data boundaries

The Assistant answers only from data you already have access to in the platform. It has no view of other users, other organizations, or anything outside your account. If you ask about company-wide data as a regular user (without admin access), the Assistant will politely refuse.

Conversation memory

The chat keeps context within a single conversation — you can follow-up without re-explaining. Between separate chat sessions the Assistant starts fresh, so no long-term memory of prior sessions.