Two operational management changes ship together in this release. The MCP audit trail records every tool invocation made through the MCP server — the authenticated user, the tool name, the input parameters, and the timestamp — providing a durable, queryable log for compliance review and agent activity analysis. The second change makes protocol service activation explicit per node: each server in the cluster declares the set of protocol contexts it runs, and the startup process activates only the declared services, replacing a previous model where all services started unless individually disabled.
MCP tool-invocation audit trail
- Per-invocation records. Each MCP tool call generates an audit entry recording the authenticated user, the tool identifier, the full input parameter set, and the invocation timestamp. The log is queryable through the standard platform reporting surface.
- Compliance and incident review. The audit trail supports both continuous monitoring — which agents are using which tools, at what frequency — and retrospective review of specific tool calls following an incident or compliance audit.
- Applies to all registered tools. Every tool registered on the MCP server is covered by the audit trail; no per-tool configuration is required to enable logging.
Per-node service declaration
- Explicit service selection. Each node declares which protocol contexts it activates — HTTP, REST, MCP, FTP, NFS — through its node configuration. The startup sequence reads this declaration before binding any network context.
- Reduced attack surface. Nodes dedicated to a single role — a REST-only API node, an MCP-only agent gateway — activate only their declared protocols rather than the full service set.
- No startup-time overrides required. Previously, disabling a service required a startup flag or environment variable. The declaration-based model makes the service set part of the node's persistent configuration, visible and version-controlled alongside other node settings.
Both changes take effect from the next server restart without application-level changes. Administrators reviewing existing deployments should verify that each node's declared service set matches its intended operational role before upgrading.