HAIEC strengthens MCP tenant-boundary verification with Tenant Isolation 2.0

A trusted MCP server can still mishandle tenant scoping. In August 2026, HAIEC documented version 2.0 of MCP Tenant Isolation, an open-source tool for examining cross-tenant exposure patterns in MCP systems. The work focuses on boundaries around paths, shared state and access scope.

The published release documentation describes support for the MCP v2 SDK, path boundary enforcement, structured scan outputs, evidence receipts and npm trusted publishing. It reports 57 deterministic rules and a suite of 203 tests for the release. Those test results describe the reported release validation, not a guarantee that unrelated MCP services are secure.

MCP Tenant Isolation gives developers a focused security check for one important layer of agentic infrastructure. HAIEC’s broader assurance model uses such evidence alongside policies, permissions, application code and observations when evaluating what an AI system could actually cause.

Model Context Protocol connections help agents discover and invoke tools, but a correctly shaped tool call does not guarantee the underlying application enforces tenant separation. Shared caches, path traversal, unscoped database reads and resource identifiers can all create cross-customer exposure in a SaaS environment.

MCP Tenant Isolation addresses the code patterns behind these risks. The tool provides deterministic source analysis for selected isolation weaknesses, including tenant-scoping omissions and access-control paths that deserve review before they become operational incidents.

The August 21, 2026 release introduced updates for MCP v2 SDK integration, stronger path handling, structured scan outputs and trusted publishing. The public repository describes a 57-rule analysis catalog. Those rules translate common tenant boundary concerns into named checks that can be run repeatedly on suitable source code.

The release’s documented test suite supports confidence in the package’s intended behavior, while remaining distinct from validating every customer integration. Tenant boundaries ultimately require correct application logic, identity propagation, authorization and runtime configuration.

A scanner finding can reveal a source-visible path that deserves engineering attention. HAIEC’s broader assurance approach can use that type of finding as one piece of evidence, alongside policy, effective permissions and observed behavior, when evaluating consequences.

Publishing this targeted tool is part of HAIEC’s developer security strategy: make high-impact checks accessible, keep detection semantics intelligible and support remediation decisions with concrete source context.