Features
Agent and entitlement discovery: Discovers AI agents in supported connected applications and shows their associated models, MCP servers, and entitlements. Shows the owner recorded for each entitlement, where available, to support approval and review.
Agent ownership: Automatically assigns owners to newly discovered unmanaged agents based on assignment rules evaluated in priority order.
Interactive agent access graph: Shows what one agent can reach, including its roles, API permissions, guardrails, knowledge bases, tools, MCP servers, models, and the other agents it works with.
Over-privileged agent flagging: Identifies agents that hold entitlements marked as privileged, helping reviewers quickly spot high-risk access during access reviews and when creating access templates.
Reusable access templates: Groups entitlements from multiple applications into a reusable set for access certification campaigns.
Scheduled certification campaigns: Schedules a review of agent access, and sets what is reviewed, which agents are included, who reviews it, how often the review repeats, and what happens to anything left undecided.
AI risk assessment: Measures each connected AI account against risk indicators mapped to the NIST framework, and calculates a risk score out of 100. The score is tracked across assessments, and each failed indicator carries a severity, the affected users, keys, or settings, and the remediation steps.
MCP gateway: Routes agent tool calls through a gateway with a single address in place of each MCP server's own, and checks every call against the control policies attached to that gateway before letting it through.
Access control policy: Allows or blocks gateway-routed tool calls based on contextual factors such as time, IP address, MCP server etc. Includes built-in policies based on OWASP Top 10 and MITRE ATLAS recommendations.
Data loss prevention (DLP) policy: Detects sensitive data in gateway-routed tool calls using built-in regex patterns for PII or custom patterns, and redacts or blocks matches based on the configured policy.
Unified dashboard: Brings together the agent and MCP server counts, guardrail coverage by platform, the enterprise AI risk score, the most used MCP servers, and recent gateway activity, and every figure leads to the records behind it.
Prebuilt reports: Provides platform reports on discovered agents, recent changes, blocked and unmanaged agents, privileged access, and associated tools, knowledge bases, and guardrails. Gateway reports cover routed tool usage, server errors, and slow calls.
Audit report: Records gateway-routed tool calls and their outcomes, along with supported technician and administrator actions in SecureAI, for later review.
Help desk delegation: Grants each technician only the part of SecureAI their work requires, through built-in roles or roles built from individual permissions. A separate workflow role of reviewer, approver, or executor decides which governance stages a technician can be assigned.