Agent identity & gateway controls — runtime security & MCP hardening
Key Questions
What recent security incidents have affected agent systems?
Incidents include Langflow ransomware, Tenet Agentjacking, LiteLLM RCE, and Claude Code secret leak.
What governance tools are available for runtime security?
Tools include Citrix MCP Gateway, Kong runtime enforcement, TBAC, Netzilo, Atlassian DLP, Databricks Unity AI Gateway, and WitnessAI.
How are identity best practices applied to MCP?
Best practices focus on task-scoped permissions and client-side runtime shifts for enterprise agents.
What new papers address agentic AI security?
The Verifiable Context Governance paper and Two-Channel Model for agentic AI security provide updated guidance.
Which frameworks support agent identity controls?
IBM × Anthropic framework and multiple gateways like Zuplo and Bifrost enable secure deployments.
Security incidents: Langflow ransomware, Tenet Agentjacking, LiteLLM RCE, Claude Code secret leak. New governance: Citrix MCP Gateway (single-pass low latency), Kong runtime enforcement, TBAC for task-scoped permissions, Netzilo runtime governance, Atlassian DLP, Databricks Unity AI Gateway, WitnessAI (99.3% true positive), Zuplo, Bifrost. New signals: Verifiable Context Governance paper, Saferagenticai MCP Server, Govern All AI Traffic gateway reference, Two-Channel Model for agentic AI security. Identity best practices for MCP deployments. Enterprise agents shifting to client-side runtime. IBM × Anthropic framework.