JAMS adds JAX agent and an MCP connector so job scheduling can be queried from coding agents

What changed: JAMS released JAX, an AI agent inside the JAMS Web Client that diagnoses failed jobs and explains fixes in plain language, and JAMS MCP, a Model Context Protocol connector that lets external coding agents (Cursor, Copilot in VS Code, Claude, Codex) query JAMS data.

Why it matters: Bringing job scheduling into the same agent-driven workflows your engineers already use removes context‑switching and lets coding agents investigate or surface operational issues without giving them elevated accounts or moving data offsite.

Try/watch: Pilot JAMS MCP with read/diagnostic queries first (no writes) so teams can validate answers and audit trails; confirm every MCP action maps to an auditable JAMS log entry before enabling any agentic changes.

PagerDuty: operators are shifting left with agent risk scores, incident memory, and self‑improving agents

What changed: PagerDuty’s Tour writeup highlights three operational moves — shift‑left risk scoring in the developer flow, using agent plugins (via protocols like MCP) to surface past incident context, and building feedback loops so agents learn from human incident playbooks.

Why it matters: For founders and operators, that means tooling now exists to catch risky vibe‑coded changes earlier and to turn post‑incident knowledge into actionable signals that coding agents can use safely, reducing the chance an agent makes dangerous autonomous decisions at 2 a.m.

Try/watch: Add automated risk checks into pull request workflows and capture incident playbooks in machine‑readable form so your agents can reference real fixes; monitor false positives and tune the scoring model to avoid blocking useful, low‑risk agent output.