kOlapsis/maintenant v1.3.4

Improve MCP server

on GitHub

A release focused on the MCP server: it fixes four write tools that silently did nothing, and adds twelve read-only tools so AI assistants can see your security posture, Kubernetes, and Docker Swarm ? bringing the server to 44 tools, all functional.

Bug fixes

MCP ? acknowledge, incident, and maintenance tools were stubs

Four MCP write tools always returned "this feature requires an extended edition" without ever calling a service ? they looked available to the AI assistant but did nothing:

• acknowledge_alert ignored the alert store entirely. Acknowledgment is a Community feature (the REST API exposes it on every edition), so the tool should always have worked. It now acknowledges an active alert, records who did it, and stops any running escalation ? mirroring POST /api/v1/alerts/{id}/acknowledge.
• create_incident, update_incident, and create_maintenance were wired to no-op managers that have no create/update methods, so they could never persist anything. They are now wired to the real status-page stores: create an incident, post a status update to its timeline, and schedule a maintenance window. These are gated to Pro, matching the REST routes.

MCP ? list_agents and the tool reference

The previously undocumented escalation, trigger, and list_agents tools are now listed in the docs, and the MCP tools reference is grouped by domain with a per-tool edition column. acknowledge_alert is correctly shown as a Community tool.

New features

MCP ? 12 read-only tools for security, Kubernetes, and Swarm

The MCP server now exposes four domains that were only reachable over REST, so an assistant can correlate vulnerabilities, cluster state, and orchestration health alongside containers and alerts. Each tool is wired to the same service the REST API uses, with Pro gating matching the REST routes:

• Security & supply chain ? get_security_insights (Community), list_cve, list_risk_scores, and get_security_posture (Pro).
• Kubernetes ? list_kubernetes_namespaces, list_kubernetes_workloads, list_kubernetes_pods (Community), and list_kubernetes_nodes (Pro). Workloads and pods accept namespace, workload, node, and status filters; agent_id selects a cluster (local for the server's own runtime).
• Docker Swarm ? get_swarm_info, list_swarm_services, list_swarm_tasks (Community), and list_swarm_nodes (Pro). get_swarm_info reports active: false when Swarm is not detected.

All twelve are read-only and carry the MCP read-only hint. Your assistant can now answer "any critical CVEs in my images?", "which workloads aren't fully ready?", or "how many Swarm replicas are running?" without leaving the editor.

Upgrade notes

No configuration changes, database migrations, or API breaking changes. Rebuild the image to pick up the new tools. MCP clients should refresh their tool list to discover the twelve new tools; the four fixed write tools keep the same names and inputs, so existing prompts continue to work ? acknowledge_alert now succeeds instead of returning an error.