AppRisingofficial publisher

mcp-scan

App : A product you use directly

Scans your MCP servers and installed skills for tool poisoning, rug pulls, and injections.

Run it before and after every community MCP installation.

GitHub · 3,169 ★

Overall score

9.3 /10

Security: nothing to flag

Utility

8.5

Reliability

10.0

Security

No caveat on record

Ease

1 click · Comfortable

Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.

What it does, concretely

MCP (Model Context Protocol) servers and AI agent skills can contain malicious instructions hidden in tool descriptions, prompt injections, or malware payloads that are hard to spot manually. mcp-scan, now integrated into Snyk as Agent Scan, addresses the need to verify components before trusting them.

The tool automatically scans agent configuration files on the local machine (Claude, Cursor, VS Code, Windsurf, Gemini CLI, Amazon Q, and others) to find MCP servers and skills. It then connects to each server, retrieves tool descriptions, and submits them to a security analysis engine that detects more than 15 distinct risk types: prompt injections, tool poisoning, tool shadowing, secret handling, suspicious downloads, and malicious code.

Agent Scan is aimed at developers and technical teams who use AI agents with community or third-party MCP servers and want to verify what they install before running it in their work environment. An enterprise offering also allows security teams to monitor the entire fleet continuously via the Snyk Evo platform.

The tool is not suited for users who have no MCP servers or skills installed, nor for those looking for a graphical interface: it works exclusively from the command line.

What it offers

·Automatic discovery of MCP servers and skills on the machine for Claude, Cursor, VS Code, Windsurf, Gemini CLI, Amazon Q, and other agents.
·Detection of more than 15 risk types: prompt injections, tool poisoning, tool shadowing, malicious code, secret handling, suspicious downloads, and private data exposure.
·Scope-based analysis (system, user, project, extension) for each supported agent.
·Explicit user consent prompt before executing each stdio MCP server in interactive mode, showing the exact command to be run.
·Scored risk report (version 0.6 and later) to help prioritize detected issues.
·Option to skip skill analysis using --no-skills to narrow the scan scope.
·Background monitoring mode with results sent to the Snyk Evo platform for security teams (paid).
·GPG-signed standalone binaries available for macOS, Linux, and Windows, with SBOM and verifiable checksums.

Our tips

Run the scan inside a Docker container or virtual machine when evaluating unverified third-party MCP configurations, as the scan actually executes the commands defined in the config files.
Use --no-skills if you want to limit the analysis to MCP servers only and reduce execution time.
Verify the integrity of the downloaded binary by checking the GPG signature of the checksums file before using it in a sensitive environment.
Pin a specific version (for example uvx snyk-agent-scan@0.5.17) if you integrate the scan into a script, as the output format is declared experimental and may change between releases.
Set the SNYK_TOKEN environment variable before running the scan: without a valid Snyk token, the analysis cannot run.

Watch out

The scan actually executes the stdio MCP server commands found in configuration files: a malicious configuration could therefore run during the analysis itself if you are not working in an isolated environment.
The output format (issue codes, field names, scores) is declared experimental and may change without notice between versions, making any automation based on raw output fragile.

In practice

Price: Free
Commercial use: Commercial use: yes
Available on: cli
Licence: Apache-2.0

Security and data

Rating: Security: nothing to flag
Your data: Local or cloud depending on use
Maintenance: Active
Last release: 20 October 2026

Links and repository checked on 25 October 2026 · added on 1 April 2025

How to install mcp-scan

1. Create an account at snyk.io and retrieve your API token from https://app.snyk.io/account. 2. Set the environment variable: export SNYK_TOKEN=your-token-here. 3. Install uv if not already available (https://docs.astral.sh/uv/getting-started/installation/). 4. Run a full machine scan: uvx snyk-agent-scan@latest. 5. To scan a specific configuration file: uvx snyk-agent-scan@latest ~/.vscode/mcp.json. 6. If you prefer a standalone binary, download it from GitHub Releases, verify the GPG signature of the checksums file, and check the binary's integrity before running it.

uvx snyk-agent-scan@latest

In the same category

Whole category Monitor →