Log inSubmit server
MCP Fortress logo

MCP Fortress

Scans MCP packages for vulnerabilities, prompt injection, tool poisoning, dependency risks, and runtime threats.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install MCP Fortress

Pick your client, then add it in one click or paste the config.

No confirmed install command yet

We only publish a config when the install details come from the project itself — its README, its docs, or a verified owner. We haven’t found those for mcp-fortress, and we’d rather show nothing than a guess you’d paste into your client.

Open setup steps on GitHub

Overview

The mcp-fortress MCP server exposes security analysis tools for scanning MCP packages and inspecting tool definitions. It checks npm and PyPI packages for vulnerabilities, suspicious code patterns, dependency CVEs, prompt injection, and tool poisoning, then produces a 0–100 risk score. Run it locally through npm or connect through Smithery, and use it when evaluating an MCP server before installation or monitoring one during operation. A local web dashboard, CLI, SQLite storage, quarantine workflow, and WebSocket telemetry support standalone use.

Use cases

  • Scan an MCP package before installing it
  • Analyze tool descriptions for prompt injection
  • Detect typosquatted or misleading tool names
  • Monitor a running MCP server
  • Quarantine suspicious packages

Key features

  • npm and PyPI vulnerability scanning
  • CVE and dependency analysis
  • Prompt injection detection
  • Tool poisoning and typosquatting detection
  • Risk scoring from 0 to 100
  • Local dashboard with monitoring and quarantine

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked 1d ago

Recent health check succeeded.

4/4 checks healthy over the last 41d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 9/13/2026 via OSV.dev

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars4Stargazers on the source repository.
Last commit6mo agoMost recent push to the default branch.
Directory activity3 viewsConfig copies, upvotes, and views on AllMCPs.

How MCP Fortress works

What mcp-fortress does

The mcp-fortress MCP server helps developers assess MCP packages and tool definitions before allowing them into an environment. Its scanner covers npm and PyPI packages, known CVEs, dependency relationships, suspicious code patterns, malicious dependencies, and license compliance issues. Results include a risk score from 0 to 100, with low, medium, and high-risk ranges.

The server also analyzes text and tool metadata for threats specific to agent workflows. It can identify instruction injection, role manipulation, attempts to extract system prompts, delimiter injection, typosquatting, misleading name or description pairs, and overly generic tool names. For example, it can flag a tool named read_fiile as a possible imitation of read_file.

How it works

The mcp-fortress MCP server can run in MCP server mode, allowing Claude Code, Cursor, or Windsurf to call its analysis tools through an MCP connection. The available tools are scan_mcp_server, analyze_prompt_injection, and detect_tool_poisoning. These tools use the requesting assistant’s existing model for AI-powered analysis, according to the project documentation.

For standalone operation, the command-line interface starts an Express API server and a React dashboard. The system stores local data in SQLite under ~/.mcp-fortress/, including the database, daemon process ID, and logs. WebSocket telemetry supports live monitoring and activity updates. The runtime features include monitoring MCP servers, quarantining suspicious packages, reviewing quarantine entries, and releasing an entry manually.

Setup and configuration

Install the package globally with npm:

Terminal
npm install -g mcp-fortress

Use mcp-fortress start to launch the standalone service. The documented defaults bind the API to localhost and use port 3001; --host, --port, --no-browser, and --daemon change startup behavior. Daemon management is available through stop, status, and logs commands.

For local MCP use, configure an MCP client to run mcp-fortress serve-mcp. The README gives a Claude Desktop configuration using the command and args fields. Smithery also provides a remote HTTP option, either through its CLI installer or a URL containing an API key and profile.

Tools and capabilities

  • Scan an npm package or a specific version with the CLI.
  • Scan a PyPI package by selecting the PyPI registry.
  • Analyze an MCP server for vulnerabilities, malicious patterns, dependency CVEs, and risk.
  • Detect prompt injection and tool poisoning through MCP tools.
  • Monitor a running server and stream activity through the dashboard.
  • List or release quarantined servers.
  • View sorted and filtered server records and detailed threat analysis in the web UI.

Limitations and notes

The documented local tier stores data locally and does not include cloud synchronization or team features. The project lists real-time MCP proxy mode, IDE configuration discovery, custom policies, CI/CD integrations, SBOM generation, and a VS Code extension as roadmap items rather than current capabilities.

Risk scores and detections are assessment outputs, not a guarantee that a package is safe. Smithery access requires an API key and profile, while the local configuration does not require an API key. The project is licensed under MIT.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about MCP Fortress

mcp-fortress is an MCP server that connects AI coding assistants to security analysis for npm and PyPI packages and MCP tools. Its main tools scan servers for vulnerabilities, analyze text for prompt injection, and detect tool poisoning such as typosquatting, misleading descriptions, and suspicious tool names.

Install it globally with npm using `npm install -g mcp-fortress`, then configure your MCP client to run `mcp-fortress serve-mcp`. The project also documents Smithery installation for a remote HTTP connection and a Claude Code plugin installation path.

Yes, mcp-fortress documents Cursor as a supported client for MCP server mode. After installing the package and configuring Cursor to connect to the MCP server, you can use its security analysis tools for server scanning, prompt injection analysis, and tool poisoning detection.

Yes, the CLI can scan PyPI packages by setting the registry to `pypi`. For example, `mcp-fortress scan flask --registry pypi` scans the named package, while npm packages can be scanned without that registry option.

It analyzes content for instruction injection attempts, role manipulation, system prompt extraction attempts, and delimiter injection. The `analyze_prompt_injection` MCP tool performs this analysis and can report suspicious text such as instructions to ignore previous directions.

Related MCP servers

All in Security All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewMCP Fortress AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/mcp-fortress?style=directory)](https://allmcps.com/mcp/mcp-fortress)
HTML Embed
<a href="https://allmcps.com/mcp/mcp-fortress"><img src="https://allmcps.com/api/badge/mcp-fortress?style=directory" alt="MCP Fortress on AllMCPs" /></a>