Skip to content

Security: subodhkc/mcp-tenant-isolation

Security

SECURITY.md

Security Policy

Scope

This policy covers the mcp-tenant-isolation npm package and its source code at https://github.com/subodhkc/mcp-tenant-isolation.

The scanner is a static analysis tool. It does not execute code from scanned projects. It parses source files and evaluates pattern-based rules.

Reporting a Vulnerability

Email: security@haiec.com

Include:

  • Description of the vulnerability
  • Steps to reproduce
  • Affected version
  • Potential impact

Response time: 48 hours for acknowledgment, 7 days for assessment.

Security Considerations

Transport (v2.0.0)

  • The MCP server uses stdio transport only. SSE transport was removed in v2.0.0.
  • The server does not bind to any network port. It communicates over stdin/stdout.
  • No network requests are made during scanning or MCP operation.

Path boundary enforcement (v2.0.0)

All filesystem operations during MCP scans are constrained to the project root configured at server startup. The PathBoundary class enforces:

  • Path traversal rejection — ../ sequences that escape the project root are rejected.
  • Symlink escape detection — symlinks that resolve outside the project root are rejected via realpath.
  • UNC/network path rejection — \\server\share paths are rejected on Windows.
  • Windows case-insensitive normalization — path comparisons are case-insensitive on Windows.
  • Absolute path containment — absolute paths outside the project root are rejected.

This boundary applies to: scan paths, config files, suppression files, baseline files, custom rulepack paths, and output file paths.

Write-tool gating (v2.0.0)

The MCP server is read-only by default. Three tools are always available:

  • scan_tenant_isolation — read-only scan
  • list_tenant_isolation_rules — read-only rule listing
  • explain_tenant_isolation_rule — read-only rule details

The suppression tool (suppress_tenant_isolation_finding) is only registered when the server is started with --allow-write-tools. Without that flag, the write tool does not exist in the MCP tool list.

Data handling

  • Suppression files (.mti-suppressions.json) are project-local and should be reviewed in code review.
  • Baseline files (.mti-baseline.json) contain finding fingerprints and file paths — no sensitive data.
  • The scanner does not collect telemetry or send scan results anywhere.

Supply chain (v2.0.0)

  • npm Trusted Publishing — packages are published via GitHub Actions OIDC, not long-lived tokens. Each publish uses a short-lived, workflow-specific credential.
  • Provenance attestations — every published package includes a signed provenance statement in the sigstore transparency log.
  • 0 npm audit vulnerabilities — dependencies are audited on every CI run.
  • Lockfile committed — CI uses npm ci for reproducible installs.

Dependency Security

  • @babel/parser — used for AST parsing, runs in Node.js process
  • commander — CLI framework
  • @modelcontextprotocol/server — MCP server transport (v2 SDK, stdio)
  • zod — schema validation for MCP tool input schemas
  • fast-glob — file discovery

All dependencies are pinned in package.json. Run npm audit regularly.

There aren't any published security advisories