1.3.0

AI Agents and Missions

2026-06-05 · #2 · closes #1

The full agent feature: a registered Agent entity, server-side Missions that match and dispatch work to it, an Agent process with a Claude sidecar, an MCP surface for the sidecar to use, and per-scope authorization with audit attribution end to end.

• Agent entity alongside Runner — registration, identity, role assignments and lifecycle, mirroring Runner's shape.
• Missions at every scope (organization, stack, namespace, module), each carrying a mission type and optional sidecar name, created through the same REST, dashboard and Terraform channels as every other resource.
• Three mission types to start — AutoDiagnose, ApprovalRecommend, SummarizeJob — dispatched from saga events on job completion, failure, cancellation and declined approval.
• Supply/demand authorization — the scope owner controls mission creation, the agent owner controls supply, and dispatch requires both.
• MCP server exposing resources, tools and prompts to the sidecar, license-gated to Enterprise.
• AgentHub (SignalR) for the agent's persistent connection, carrying dispatch, heartbeat, log streaming and run lifecycle.
• Audit attribution — an agent_id claim threads agent identity through every REST call it makes.