529 posts tagged with "AI agents"
AI agents and autonomous systems
Your Sandbox Is an L2 Guest: What E2B's Nested Firecracker Really Costs vs Bare Metal
E2B's sandboxes run as L2 guests on rented GCP VMs. Fingerprint data, latency benchmarks, and a line-by-line bill recompute show where renting wins — and the roughly 300 sandbox-hour crossover where owned hardware takes over.
Firecracker's First Escape CVEs: What CVE-2026-5747 and CVE-2026-1386 Mean for MicroVM Sandboxes
Firecracker's zero-escape-CVE record broke twice in 2026: an out-of-bounds write in the virtio-pci transport and a symlink overwrite in the jailer. What both bugs mean for teams running AI-agent sandboxes on microVMs, and the patching checklist that keeps the isolation story intact.
Google Just Normalized the Governed Agent Endpoint: What Its Home and UN Data Commons MCP Servers Mean for Your Self-Hosted PaaS Roadmap
Google shipped first-party MCP servers for Home and UN data in a single week. Here is the six-part governance bar they normalized, and what it demands from any self-hosted platform shipping an agent endpoint.
Someone Already Wrote the Hetzner MCP Server: What a Live Changelog Audit Teaches Deploy-From-Chat
A community MCP server mapping the Hetzner Cloud, Storage Box, and Robot APIs survived six provider changelog deltas in three months with zero broken tools — thanks to a published endpoint audit, pass-through writes, and an eval that expects deprecation. Here is what that discipline teaches any deploy-from-chat roadmap, and the tenant-scoping and audit-logging layer a PaaS still has to build itself.
Kubernetes v1.37 Node Lifecycle Conditions: Machine-Readable Node State for Agent-Driven Fleet Ops
Kubernetes v1.37 adds five Node Lifecycle Conditions — DrainInProgress, Drained, MaintenancePlanned, MaintenanceInProgress, and GracefulNodeShutdownInProgress — so nodes report intent, not just readiness. Here is who should publish each one on a Cluster API fleet, how to rewrite NotReady paging, and the agent policy that turns machine-readable state into safe machine operators.
One Scheduler Per Cluster Isn't a Fleet Strategy: How KubeStellar, Cluster API, and Hive Split the Multi-Cluster Job
kube-scheduler goes blind the day you add a second cluster. How Cluster API provisioning, KubeStellar placement policy, and hive agent orchestration split the multi-cluster job — and when placement belongs in your platform versus tenant BYO.
Laravel MCP 1.0 Ships: Your Laravel App Is Now an Agent Tool — What Changes on the Deploy Surface
Laravel MCP 1.0 adopts the stateless MCP 2026-07-28 spec with searchable tool catalogs, cache hints, and mandatory OAuth PKCE — plus header validation that 400s old clients. Here is the upgrade checklist and what per-app MCP servers mean for anyone operating Laravel apps.
Your MCP Deploy Tool's dry_run Flag Is a Suggestion, Not a Lock
MCP tool annotations like readOnlyHint are advisory hints for the client, not enforcement — nothing in the protocol stops a caller from overriding dry_run, tenant_id, or target_environment. A three-line server-side pattern (derive from the session, reject client values, log the attempt) and where host-authority receipts belong.
MCP Just Got Its Biggest Update Ever: What Stateless Servers Mean for Self-Hosting Deploy-from-Chat
The 2026-07-28 MCP spec killed sessions, sticky routing, and handshake affinity — September coverage calls it the biggest update since launch. For self-hosters, that means an infrastructure MCP server is now an ordinary HTTP service behind the same load balancer as everything else.