Skip to main content

KAI (Korweave Augmented Insights)

KAI is Kscope's GenAI-powered assistant. Ask security and operational questions in plain English and get instant, context-aware answers drawn from your context graph -- no scripting, SQL, or manual correlation needed.

KAI uses Retrieval-Augmented Generation (RAG) against your organization's actual asset graph and security data. All queries stay within your security controls.

What KAI Can Answer​

QuestionWhy it matters
Which S3 buckets are publicly accessible?Detect misconfigured storage that may lead to data leaks
Show me business-critical apps with critical CVEs unpatched for 30+ daysPrioritize remediation by business impact
List all internet-facing assets with high-severity vulnerabilitiesFocus defenses on the externally visible attack surface
Are any Kubernetes clusters running outdated images with known CVEs?Ensure container environments aren't vulnerable
Which third-party dependencies have exploitable vulnerabilities?Reduce software supply chain risk
Do any IAM roles have excessive permissions in production?Mitigate privilege escalation risk
What are the top 5 exploitable misconfigurations in our cloud estate?Prioritize fixes that reduce the biggest risks

Layout​

KAI Interface

  • Session History
  • Sample Questions
  • Input Bar

LLM + RAG

Context Graph

Response

  • Chart
  • Graph
  • Text
  • Table
  • Sample Questions -- Pre-defined clickable queries to get started immediately
  • Input Bar -- Type any question in natural language
  • Session History -- Browse and continue past conversations (sidebar)
  • Send Button -- Submit your query

Response Types​

KAI returns answers in the format that best fits the data:

Chart response -- "Show me EC2 Instance State Distribution as a pie chart"

Graph response -- "Show me S3 Buckets"

Text response -- "What are the names of S3 Buckets"

Saving Responses​

Save any chart or graph to a Custom Analyzer as a Sightline:

  1. Click Create Sightline on the response
  2. Fill in the name and details
  3. Access it later from your Custom Analyzer

Tips​

  • Be specific -- "What are the properties of node A?" works better than "What is the node data?"
  • Start with sample questions if you're unsure how to phrase a query
  • Use follow-ups -- KAI refines answers based on conversation context
  • Save important findings -- Use Sightlines to organize insights for later

K-MCP (Kscope Model Context Protocol)​

K-MCP extends KAI's capabilities to external AI tools and IDE assistants through the Model Context Protocol (MCP). It exposes your Kscope context graph as an MCP server, allowing AI coding assistants and security tools to query your security posture directly.

What K-MCP Enables​

CapabilityDescription
IDE integrationQuery your security posture from AI assistants in VS Code, JetBrains, or other MCP-compatible editors
Context-aware developmentGet real-time security context about the services you're working on without leaving your editor
Automated triageFeed security findings into AI workflows that can prioritize and suggest remediations
Custom toolingBuild MCP clients that integrate Kscope data into your internal security automation

Connection Details​

To view your K-MCP connection details, click the MCP icon next to New Conversation on the KAI page. The dialog shows:

AttributeDescription
K-MCP URLThe SSE endpoint for your MCP client to connect to (e.g., https://api.<tenant>.kscope.ai/k-mcp/sse)
Domain URLThe authentication domain for OAuth token exchange
ScopeGranted permissions -- k-mcp/read, ke/view_analyzers, ke/acknowledge_alert, ke/get_account
Client IDThe OAuth client ID for authenticating your MCP client
Client SecretThe OAuth client secret -- treat this as a password
Grant Typeclient_credentials -- use OAuth 2.0 client credentials flow

Connecting an MCP Client​

  1. Open the K-MCP connection details dialog from the KAI page
  2. Copy the K-MCP URL, Client ID, and Client Secret
  3. Configure your MCP-compatible client with these credentials using the client_credentials OAuth flow against the Domain URL
  4. Connect to the K-MCP URL (SSE transport)
  5. Start querying your security posture from your preferred tool

Example: Claude Desktop / Claude Code

Add to your MCP configuration:

{
  "mcpServers": {
    "kscope": {
      "url": "https://api.<tenant>.kscope.ai/k-mcp/sse",
      "headers": {
        "Authorization": "Bearer <access_token>"
      }
    }
  }
}