The Capsule Security platform helps you discover, monitor, and secure AI agents across your organization.
Connect with leading AI platforms:
- Microsoft: Copilot Studio, Power Automate
- Salesforce: Agentforce
- OpenAI: ChatGPT Enterprise
- ServiceNow: ServiceNow AI Agents
- IDEs: Cursor, Claude Code
- AWS: Bedrock
- Azure: Azure AI Foundry
- Google Cloud: Vertex AI
- Automation: Zapier, UiPath, n8n
- Databricks: Agent Bricks
- Enterprise: SAP Joule Agents, Box AI Studio, Workato Genies
Once you integrate a platform, Capsule automatically finds:
- All AI agents in your environment
- What tools and data sources each agent can access
- Who owns each agent
- Where agents are deployed
- How agents are being used
Browse all discovered agents with:
- Visual severity indicators showing risk levels
- Platform icons for quick identification
- Owner information
- Last activity timestamps
- Quick access to detailed information
Use filters to find specific agents by platform, owner, severity, or other attributes.
- Agent name and description
- Platform and environment
- Owner name and contact information
- Creation date
- Unique identifiers
Agents can have three accessibility levels:
- PUBLIC: Available to anyone
- TENANT: Available to users in your organization
- LIMITED: Available to specific users only
- Issue severity (CRITICAL, HIGH, MEDIUM, LOW)
- Number of open issues
- Issue categories affecting the agent
View what the agent can access:
- Database connections
- API integrations
- File systems
- Email services
- External systems
See how data moves through your agent:
- Connected data sources
- Tools the agent uses
- Flow direction between components
- Related security issues
Track agent usage:
- Total sessions
- Recent activity
- Usage patterns over time
- Session details and audit logs
Issues are organized by risk type:
- SUPPLY_CHAIN: Risks from third-party dependencies
- DATA: Data exposure or leakage
- COMPLIANCE: Policy violations
- INTEGRITY: Data quality concerns
- ACCESS: Unauthorized access risks
- MISC: Other security concerns
- CRITICAL: Fix immediately
- HIGH: Fix within 24 hours
- MEDIUM: Fix within 1 week
- LOW: Fix in next maintenance window
- OPEN: Needs attention
- RESOLVED: Fixed and verified
- IGNORED: Accepted risk
- View all issues or filter by severity, category, or status
- Assign issues to team members
- Follow remediation steps
- Mark steps as completed
- Update issue status when resolved
Track all agent activity:
- Session Started: When someone starts using the agent
- Tool Call: When the agent uses a tool or service
- Message: User interactions with the agent
- Error: Problems that occurred
Filter audit logs by date range and review issues detected during sessions.
- Assign an owner to every agent
- Owners are responsible for security and compliance
- Review ownership regularly
- Use LIMITED accessibility when possible
- Grant only necessary permissions
- Review agent access regularly
- Only enable tools the agent needs
- Remove unused connections
- Review tool access regularly
- Address CRITICAL and HIGH severity issues first
- Document how you resolved issues
- Look for patterns to prevent future issues
- Review agent activity regularly
- Set up alerts for unusual behavior
- Track usage trends
- Check that your platform integration is enabled
- Verify platform permissions
- Wait a few minutes for the next sync
- Ensure platform API permissions are correct
- Check your network connection
- Review error messages in the integration settings
- Review the issue details and reasoning
- Check the category and remediation steps
- Contact support if ratings seem incorrect