mcpField
R
reachvuln

reachvuln

Local
github.com/CloudCorpRecordsSecurity0Updated 1w ago

About reachvuln

reachvuln is a local scanner-report ranking tool for developers who need to prioritize vulnerabilities by code reachability. It analyzes supplied imports and entry points, then applies explicit severity weights and reachability multipliers.

Key features

  1. 01Build reachability from an import graph
  2. 02Match vulnerable package exports
  3. 03Classify findings by reachability
  4. 04Rank findings with fixed scoring rules
  5. 05Validate inputs and report counts

Use cases

  1. 01Prioritize scanner findings for remediation
  2. 02Identify findings reachable from application entry points
  3. 03Separate unused packages from reachable vulnerabilities
  4. 04Rank vulnerabilities using supplied module data

MCP support

Exposes the rank_vulns_by_reach tool to MCP clients. It accepts vulnerability findings, module imports, and entry points, then returns ranked results and reachability counts.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"cloudcorprecords-reachvuln": {
"command": "npx",
"args": [
"-y",
"@reneturcios/reachvuln"
]
}
}
}
Run commandnpx -y @reneturcios/reachvuln

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npm@reneturcios/reachvuln0.1.0
ocighcr.io/cloudcorprecords/reachvuln:0.1.0

Tools

1

Saved tool list · updated 1w ago

ParameterTypeRequiredDescription
modulesarrayYesModules and their imports: {path, imports}
entry_pointsarrayYesModule paths that are entry points
vulnerabilitiesarrayYesFindings: {id, package, severity, vulnerable_exports}

Details

From the project README

reachvuln

A scanner report with two hundred findings is not a work list. Most of those packages may never be loaded by your code at all, and of the ones that are, the vulnerable export may never be the one you import. reachvuln reorders the report around that question: given the findings, your modules and their imports, and your entry points, which findings can actually be reached?

Reachability is computed over the import graph you supply. An import whose package equals another module's path is a local edge; anything else is an external package import. The reachable modules are the entry points plus everything reached from them by following local edges — breadth-first, and safe around cycles, so a circular import chain neither loops forever nor hides what sits behind it.

A finding is reachable when some reachable module imports its package and either the finding lists no vulnerable_exports (any import of the package counts) or that import's names share at least one entry with vulnerable_exports. It is imported_unreachable when some module imports the package but the reachable condition does not hold — the package is in the tree, yet no reachable module imports it in a way that touches the vulnerable exports. It is not_imported when no module imports the package at all. For a reachable finding, reached_via lists the sorted distinct paths of the reachable modules whose imports satisfy the condition; otherwise it is empty.

Ranking is pinned arithmetic, not a heuristic: severity weighs critical 40, high 30, medium 20, and low 10, and the weight is multiplied by 3 for reachable findings, by 1 for imported-but-unreachable ones, and by 0 for not-imported ones. Findings order by rank score descending, then severity weight descending, then id ascending, and the result also carries counts for the three classes.

Validation is strict and names the offending path: a missing or badly typed vulnerabilities, modules, or entry_points; a finding with a missing, empty, or duplicate id, a missing or empty package, a severity outside its enum, or a badly typed vulnerable_exports; a module with a missing, empty, or duplicate path, a missing imports, or an import with a bad package or names; and an entry point that is not one of the module paths. Note that entries in an import's names may be empty strings — they simply never match a vulnerable export, which is itself required to be non-empty. Unknown extra fields are ignored, and no partial result is produced on invalid input. The computation is pure and local: no network, no configuration, and no repository is read.

CLI

npx @reneturcios/reachvuln rank --input args.json
npx @reneturcios/reachvuln --version

args.json contains the full arguments object:

{
  "vulnerabilities": [
    {
      "id": "CVE-1",
      "package": "lodash",
      "severity": "high",
      "vulnerable_exports": ["merge"]
    }
  ],
  "modules": [
    {
      "path": "src/app.ts",
      "imports": [{ "package": "lodash", "names": ["merge"] }]
    }
  ],
  "entry_points": ["src/app.ts"]
}

The result is printed as JSON:

{
  "ranked": [
    {
      "id": "CVE-1",
      "package": "lodash",
      "severity": "high",
      "reachability": "reachable",
      "rank_score": 90,
      "reached_via": ["src/app.ts"]
    }
  ],
  "counts": {
    "reachable": 1,
    "imported_unreachable": 0,
    "not_imported": 0
  }
}

Exit codes: 0 on success, 1 on a runtime error (including an unreadable or invalid input file, or invalid input), 2 on a usage error.

MCP server

npx -y @reneturcios/reachvuln mcp

MCP client config:

{
  "mcpServers": {
    "reachvuln": {
      "command": "npx",
      "args": ["-y", "@reneturcios/reachvuln", "mcp"]
    }
  }
}
Tools
  • rank_vulns_by_reach — Rank vulnerability findings by whether the vulnerable package and exports are reachable from your entry points.

    • vulnerabilities (array, required, may be empty) — findings, each {id, package, severity, vulnerable_exports}; id is a non-empty unique string, package a non-empty string, severity one of critical, high, medium, low, and vulnerable_exports an optional array of non-empty strings (default [])
    • modules (array, required, may be empty) — modules, each {path, imports}; path is a non-empty unique string and imports an array of {package, names}, where package is a non-empty string and names a required array of strings (entries may be empty)
    • entry_points (array, required, may be empty) — module paths that are entry points; every entry must equal some module's path

    Returns {ranked, counts}, where each ranked entry is {id, package, severity, reachability, rank_score, reached_via} in the pinned order and counts is {reachable, imported_unreachable, not_imported}.

Docker

docker run -i ghcr.io/cloudcorprecords/reachvuln
docker run --rm -v "$PWD:/w" ghcr.io/cloudcorprecords/reachvuln rank --input /w/args.json

Development

npm install
npm run build   # tsc -> dist/
npm test        # tsx --test test/*.test.ts (pure local, no network)
npm run dev     # run the CLI from source

License

MIT — Copyright (c) 2026 Rene Turcios

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.