@sylphx/pdf-reader-mcp 2.3.0 → 2.4.0

package/README.md

@@ -171,6 +171,10 @@ Add to Cline's MCP settings:
 1. Go to **Settings** → **AI** → **Manage MCP Servers** → **Add**
 2. Command: `npx`, Args: `@sylphx/pdf-reader-mcp`
 
+### Ontheia
+
+Add the server in **Settings** → **MCP Servers** → **Add Server** with command `npx` and args `@sylphx/pdf-reader-mcp`. See [Ontheia's compatible MCP servers](https://docs.ontheia.ai/getting-started/03_compatible-mcp-servers/) for the full list.
+
 ### Smithery (One-click)
 
 ```bash
@@ -538,6 +542,55 @@ Response Order:
 
 ---
 
+## 🔒 Security & Sandboxing
+
+By default the server can read any local file the host process can access and fetch any HTTP(S) URL. When running outside a sandbox you should restrict it to a specific working set.
+
+### Restricting filesystem access
+
+Use `--allow-dir` (repeatable) or the `MCP_PDF_ALLOWED_DIRS` env var (`:` or `,` separated). Once set, all `path` sources must resolve inside one of the allowed directories — relative paths, absolute paths, and `..` traversal are all checked after resolution.
+
+```bash
+# CLI flags
+npx @sylphx/pdf-reader-mcp --allow-dir=/srv/pdfs --allow-dir=/data/reports
+
+# Environment
+MCP_PDF_ALLOWED_DIRS="/srv/pdfs:/data/reports" npx @sylphx/pdf-reader-mcp
+```
+
+```json
+{
+  "mcpServers": {
+    "pdf-reader": {
+      "command": "npx",
+      "args": ["@sylphx/pdf-reader-mcp", "--allow-dir=/srv/pdfs"]
+    }
+  }
+}
+```
+
+### Disabling or restricting HTTP
+
+```bash
+# Block all URL sources
+npx @sylphx/pdf-reader-mcp --no-http
+MCP_PDF_ALLOW_HTTP=false npx @sylphx/pdf-reader-mcp
+
+# Allowlist hosts (everything else rejected)
+npx @sylphx/pdf-reader-mcp --allow-host=cdn.example.com --allow-host=files.internal
+MCP_PDF_ALLOWED_HOSTS="cdn.example.com,files.internal" npx @sylphx/pdf-reader-mcp
+```
+
+| Setting | CLI flag | Environment variable | Default |
+|---------|----------|----------------------|---------|
+| Filesystem allowlist | `--allow-dir=<path>` (repeatable) | `MCP_PDF_ALLOWED_DIRS` (`:` or `,` separated) | unrestricted |
+| Disable HTTP | `--no-http` | `MCP_PDF_ALLOW_HTTP=false` | enabled |
+| HTTP host allowlist | `--allow-host=<host>` (repeatable) | `MCP_PDF_ALLOWED_HOSTS` (`,` separated) | any host |
+
+Denied requests fail fast with an `Access denied` error before any disk read or network call.
+
+---
+
 ## 🔧 Troubleshooting
 
 ### "Absolute paths are not allowed"

package/dist/index.js

@@ -319,6 +319,88 @@ import fs from "node:fs/promises";
 import { createRequire } from "node:module";
 import { getDocument } from "pdfjs-dist/legacy/build/pdf.mjs";
 
+// src/utils/config.ts
+import path from "node:path";
+var splitList = (value, separators) => value.split(separators).map((s) => s.trim()).filter((s) => s.length > 0);
+var parseDirs = (values) => values.map((dir) => path.resolve(path.normalize(dir)));
+var parseBool = (value, fallback) => {
+  if (value === undefined)
+    return fallback;
+  const v = value.trim().toLowerCase();
+  if (v === "false" || v === "0" || v === "no" || v === "off")
+    return false;
+  if (v === "true" || v === "1" || v === "yes" || v === "on")
+    return true;
+  return fallback;
+};
+var parseCliFlags = (argv) => {
+  const dirs = [];
+  const hosts = [];
+  let noHttp = false;
+  for (const arg of argv) {
+    if (arg.startsWith("--allow-dir=")) {
+      dirs.push(arg.slice("--allow-dir=".length));
+    } else if (arg.startsWith("--allow-host=")) {
+      hosts.push(arg.slice("--allow-host=".length).toLowerCase());
+    } else if (arg === "--no-http") {
+      noHttp = true;
+    }
+  }
+  return { dirs, hosts, noHttp };
+};
+var envList = (raw, separators, transform = (v) => v) => raw ? splitList(raw, separators).map(transform) : [];
+var readSecurityConfig = (argv = process.argv.slice(2), env = process.env) => {
+  const cli = parseCliFlags(argv);
+  const envDirs = envList(env["MCP_PDF_ALLOWED_DIRS"], /[:,]/);
+  const envHosts = envList(env["MCP_PDF_ALLOWED_HOSTS"], /,/, (h) => h.toLowerCase());
+  const mergedDirs = [...cli.dirs, ...envDirs];
+  const mergedHosts = [...cli.hosts, ...envHosts];
+  return {
+    allowedDirs: mergedDirs.length > 0 ? parseDirs(mergedDirs) : null,
+    allowHttp: cli.noHttp ? false : parseBool(env["MCP_PDF_ALLOW_HTTP"], true),
+    allowedHosts: mergedHosts.length > 0 ? mergedHosts : null
+  };
+};
+var cached = null;
+var getSecurityConfig = () => {
+  if (cached === null) {
+    cached = readSecurityConfig();
+  }
+  return cached;
+};
+var isPathAllowed = (absPath, allowedDirs) => {
+  if (allowedDirs === null)
+    return true;
+  if (allowedDirs.length === 0)
+    return false;
+  const normalized = path.resolve(absPath);
+  return allowedDirs.some((dir) => {
+    const rel = path.relative(dir, normalized);
+    if (rel === "")
+      return true;
+    if (rel.startsWith(".."))
+      return false;
+    if (path.isAbsolute(rel))
+      return false;
+    return true;
+  });
+};
+var isUrlAllowed = (urlString, config) => {
+  if (!config.allowHttp)
+    return false;
+  let parsed;
+  try {
+    parsed = new URL(urlString);
+  } catch {
+    return false;
+  }
+  if (parsed.protocol !== "http:" && parsed.protocol !== "https:")
+    return false;
+  if (config.allowedHosts === null)
+    return true;
+  return config.allowedHosts.includes(parsed.hostname.toLowerCase());
+};
+
 // src/utils/errors.ts
 class PdfError extends Error {
   code;
@@ -330,20 +412,29 @@ class PdfError extends Error {
 }
 
 // src/utils/pathUtils.ts
-import path from "node:path";
+import path2 from "node:path";
 var PROJECT_ROOT = process.cwd();
 var resolvePath = (userPath) => {
   if (typeof userPath !== "string") {
     throw new PdfError(-32602 /* InvalidParams */, "Path must be a string.");
   }
-  const normalizedUserPath = path.normalize(userPath);
-  return path.isAbsolute(normalizedUserPath) ? normalizedUserPath : path.resolve(PROJECT_ROOT, normalizedUserPath);
+  const normalizedUserPath = path2.normalize(userPath);
+  const resolved = path2.isAbsolute(normalizedUserPath) ? normalizedUserPath : path2.resolve(PROJECT_ROOT, normalizedUserPath);
+  const { allowedDirs } = getSecurityConfig();
+  if (!isPathAllowed(resolved, allowedDirs)) {
+    throw new PdfError(-32600 /* InvalidRequest */, `Access denied: path '${userPath}' is outside the allowed directories.`);
+  }
+  return resolved;
 };
 
 // src/pdf/loader.ts
 var logger3 = createLogger("Loader");
 var require2 = createRequire(import.meta.url);
-var CMAP_URL = require2.resolve("pdfjs-dist/package.json").replace("package.json", "cmaps/");
+var PDFJS_ROOT = require2.resolve("pdfjs-dist/package.json").replace("package.json", "");
+var CMAP_URL = `${PDFJS_ROOT}cmaps/`;
+var STANDARD_FONT_DATA_URL = `${PDFJS_ROOT}standard_fonts/`;
+var WASM_URL = `${PDFJS_ROOT}wasm/`;
+var ICC_URL = `${PDFJS_ROOT}iccs/`;
 var MAX_PDF_SIZE = 100 * 1024 * 1024;
 var loadPdfDocument = async (source, sourceDescription) => {
   let pdfDataSource;
@@ -356,6 +447,11 @@ var loadPdfDocument = async (source, sourceDescription) => {
       }
       pdfDataSource = new Uint8Array(buffer);
     } else if (source.url) {
+      const config = getSecurityConfig();
+      if (!isUrlAllowed(source.url, config)) {
+        const reason = config.allowHttp ? `host is not in the allowed list` : `HTTP access is disabled`;
+        throw new PdfError(-32600 /* InvalidRequest */, `Access denied: URL '${source.url}' rejected (${reason}).`);
+      }
       pdfDataSource = { url: source.url };
     } else {
       throw new PdfError(-32602 /* InvalidParams */, `Source ${sourceDescription} missing 'path' or 'url'.`);
@@ -377,7 +473,10 @@ var loadPdfDocument = async (source, sourceDescription) => {
   const loadingTask = getDocument({
     ...documentParams,
     cMapUrl: CMAP_URL,
-    cMapPacked: true
+    cMapPacked: true,
+    standardFontDataUrl: STANDARD_FONT_DATA_URL,
+    wasmUrl: WASM_URL,
+    iccUrl: ICC_URL
   });
   try {
     return await loadingTask.promise;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@sylphx/pdf-reader-mcp",
-  "version": "2.3.0",
+  "version": "2.4.0",
   "description": "An MCP server providing tools to read PDF files.",
   "type": "module",
   "bin": {
@@ -69,29 +69,32 @@
     "prepare": "node_modules/.bin/lefthook install || true"
   },
   "dependencies": {
-    "@sylphx/mcp-server-sdk": "^2.1.0",
+    "@sylphx/mcp-server-sdk": "^2.1.1",
     "@sylphx/vex": "^0.1.11",
-    "glob": "^13.0.0",
-    "pdfjs-dist": "^5.4.449",
+    "glob": "^13.0.6",
+    "pdfjs-dist": "^5.6.205",
     "pngjs": "^7.0.0"
   },
   "overrides": {
     "esbuild": "^0.25.0",
-    "preact": "^10.28.2"
+    "preact": "^10.28.2",
+    "defu": "^6.1.7",
+    "rollup": "^4.60.2",
+    "vite": "^6.4.2"
   },
   "devDependencies": {
-    "@biomejs/biome": "^2.3.8",
+    "@biomejs/biome": "^2.4.12",
     "@sylphx/biome-config": "^0.4.1",
     "@sylphx/bump": "^1.6.1",
-    "@sylphx/doctor": "^1.32.1",
+    "@sylphx/doctor": "^1.34.0",
     "@sylphx/tsconfig": "^0.3.1",
     "@types/glob": "^8.1.0",
-    "@types/node": "^25.0.3",
+    "@types/node": "^25.6.0",
     "@types/pngjs": "^6.0.5",
     "bunup": "0.16.10",
-    "lefthook": "^2.0.7",
-    "typedoc": "^0.28.15",
-    "typedoc-plugin-markdown": "^4.9.0",
+    "lefthook": "^2.1.6",
+    "typedoc": "^0.28.19",
+    "typedoc-plugin-markdown": "^4.11.0",
     "typescript": "^5.9.3",
     "vitepress": "^1.6.4"
   },