PortunusMCP Sentinel

Build-time security scanning for MCP servers.

PortunusMCP Sentinel finds security problems in local MCP server repositories before deployment. It combines deterministic Python and TypeScript checks, GPT-5.6 semantic review, Docker-isolated runtime probes for Python targets, OWASP Agentic Top 10 mappings, and SARIF output for GitHub code scanning.

The rules-only commands below require 1.3.0 or newer; see installation and upgrades.

Start without credentials or Docker

pipx install portunusmcp-sentinel
cd your-mcp-server
sentinel scan . --rules-only

sentinel init is optional and generates only the permissions sidecar. Rules-only performs no model/cache/network/Docker operations or target execution. Installation and Action upload have separate network requirements.

Exit 0 and exit 1 both mean analysis completed. Exit 1 means at least one finding reached the configured failure threshold.

Choose an analysis tier

Tier	Command	Requirements
Rules-only	sentinel scan . --rules-only	Sentinel only
Static + GPT review	sentinel scan . --static-only	OPENAI_API_KEY
Full dynamic proof	sentinel scan .	OPENAI_API_KEY, Docker, Python target

Start with installation, then review configuration and the rule catalog. For CI, use the GitHub Action.

Measured effectiveness

The accepted independent benchmark records 45 labeled inputs across five repositories. Both static tiers completed 32 inputs and detected none of the 14 vulnerable inputs whose scans completed. Review captures are complete; execution failures and unsupported runtime scope remain explicit. This baseline guides further work and does not establish broad detection accuracy.

PortunusMCP family

Sentinel is the build-time security plane of PortunusMCP. The PortunusMCP Gateway is the separate runtime enforcement plane, and PortunusMCP Identity is the separate short-lived credential-broker plane. Sentinel does not require either project.