Blog
Field notes on governed agents
Practical writing on letting AI agents operate real infrastructure without giving up control, auditability, or sleep.
AI Agent Observability: What to Log When Agents Act
AI agent observability means tracing what an agent did to real systems, not just what the model said. What to log, and why it differs from LLM observability.
MCP Server Security: The Risks and the Fixes
MCP servers give agents real tools, and real reach. Here are the concrete security risks of MCP servers and the fixes that actually hold.
AI Agent Guardrails That Actually Work
Prompt instructions aren't guardrails. Real guardrails are enforced outside the model: typed capabilities, risk policies, approvals, audit, and rollback.
What Is AI Agent Governance? A Practical Guide
AI agent governance is the control layer that decides what autonomous agents can do in real systems: identity, approvals, audit, and rollback. A practical guide.
How to Give Claude Code Safe AWS Access
Pasting AWS keys into an agent's environment is the new shadow IT. Safer patterns for Claude Code and any coding agent, from scoped roles to a governed broker.