AI Agents
Designing agents that act, not just answer
Why action-grounded agents — armed with policies, telemetry and rollback — outperform chat-only copilots in production operations.
Aarav Mehta · Principal Engineer• 8 min read•Feb 12, 2026
MetaxuOne Blog
Ideas on autonomous operations.
Field notes from our engineers, product leads and customers — on AI agents, observability, platform design and the work of running modern enterprises.
Featured
Latest posts
Observability
From signals to context: rebuilding the operations graph
How MetaxuOne stitches metrics, traces, logs and topology into a single context model SRE teams can reason about.
Sara Iyer 11 min
Platform
A field guide to safe autonomous remediation
Guardrails, audit trails and human-in-the-loop patterns that make self-healing trustworthy in regulated environments.
Devansh Kapoor 9 min
Engineering
Inside our event pipeline: 12M events/sec, single-digit ms p99
The architecture behind MetaxuOne's ingestion plane — partitioning, backpressure, and how we cut tail latency.
Nikhil Rao 14 min
Customer Stories
How a global bank cut MTTR by 63% in six weeks
A walkthrough of the rollout, the wins, and the lessons learned moving from alert fatigue to agent-led triage.
Priya Natarajan 6 min
AI Agents
Agent memory: short, long, and operational
A practical taxonomy of memory for agents that operate on living systems — and how to keep it bounded.
Aarav Mehta 7 min
Observability
Why dashboards are not enough anymore
The shift from passive visualization to active, conversational operations driven by AI assistants.
Sara Iyer 5 min
Get the next post in your inbox.
One concise note every other week. Engineering deep-dives, operational playbooks, and the occasional opinion.