Why shallow package checks miss risk
MCP servers commonly depend on SDKs, web frameworks, transport libraries, and utility packages through several layers of transitive dependencies. A direct lookup of the package name can return no findings while the installed tree still contains a version associated with a public advisory.
The Bindfort scan approach inventories resolved versions, records the evidence source, and keeps package presence separate from runtime reachability. A finding is a review input, not automatic proof that a vulnerability is exploitable in a specific deployment.
- Inventory the exact package manager, version, and resolved dependency tree.
- Compare installed components with public vulnerability sources.
- Separate dependency presence from confirmed reachable behavior.
- Preserve a reproducible summary for remediation and follow-up scans.
A scanner should support a decision
The useful output is not a long vulnerability list. Security teams need to know which server is affected, which installed component produced the finding, whether the affected path is reachable, and what action should follow.
Bindfort’s scanner-only CLI is publicly downloadable with no account or API key. It produces local table or JSON findings; continuous registry monitoring, reachability analysis, and automatic production blocking are product-direction items, not claims about the public scanner.
Product status
Verified today and clearly separated from roadmap
- Recursive installed-tree inventory for local npm, Go, and Python project inputs.
- Public-advisory matching with evidence and version context.
- Reproducible table and JSON findings with configurable CI exit thresholds.
- Clear distinction between package presence and runtime reachability.
- Continuous production inventory and scheduled re-scanning.
- Exception workflows, ownership routing, and remediation tracking.
- Automatic policy response for verified high-risk conditions.
Frequently asked questions
Clear answers for evaluation
What does an MCP security scanner inspect?
A useful scanner inventories the MCP server, its resolved dependencies, known advisories, tool exposure, configuration boundaries, and evidence needed for review.
Does a vulnerability finding prove exploitation?
No. A dependency finding establishes presence. Reachability and exploitability require separate validation in the actual runtime context.
Can Bindfort scan a private repository?
Yes. Run the scanner locally against the checked-out project directory. Source files and lockfiles stay on the machine; normalized package coordinates are sent to the public OSV API for advisory matching. Do not include secrets in support requests.