Snyk Agent Scan: MCP vendor profile
Snyk Agent Scan, the former Invariant Labs mcp-scan, discovers agent configurations on a machine and scans their MCP servers, tools, prompts, resources and skills for prompt injection and other risks, as documented by the vendor. Agent Scan validates components with local checks and the Agent Scan API, sending agent application details, MCP server configurations and signatures for analysis (github.com).
At a glance
| Listed as | Snyk Agent Scan (formerly Invariant Labs mcp-scan) |
|---|---|
| Type | Security scanners |
| Operator | Snyk |
| Home | github.com |
| Price | price not read |
| Read | 5 November 2026 |
What it does for this layer
Running Agent Scan requires a Snyk API token set as an environment variable before any scan (github.com). Snyk does not publish an npm package for Agent Scan; it installs with uvx (for example uvx --python 3.13 snyk-agent-scan@latest scan mcp.json) or as a standalone binary (github.com). Agent Scan v0.5.x with issue-code output is planned for deprecation; v0.6 and later report risk-based output, and Snyk calls the raw CLI output experimental (github.com).
Agent Scan accepts --mcp-oauth-tokens-path to scan remote MCP servers protected by OAuth (github.com). Snyk keeps Agent Scan open source under Apache-2.0; its repository counted about 3.2k stars on 5 Nov 2026 (github.com, observed 5 November 2026). The former invariantlabs-ai/mcp-scan repository URL redirected to github.com/snyk/agent-scan on 5 Nov 2026 (github.com, observed 5 November 2026).
Snyk describes Agent Scan as a "Security scanner for AI agents, MCP servers and agent skills" that looks for prompt injections and vulnerabilities (github.com). Agent Scan scans MCP servers, tools, prompts, resources and skills and discovers configurations for Claude Code and Desktop, Cursor, Gemini CLI and Windsurf (github.com).
Products and plans
| Product | What it does | Price |
|---|---|---|
| Agent Scan | CLI scanner for agent configs, MCP servers and skills, backed by the Snyk analysis API. | price not read |
Standards
| Standard | Part of it | Status | What the page says | Source |
|---|---|---|---|---|
| Agent Skills | skills scanning | Supports | Agent Scan 0.4 added scanning of agent skills. | github.com |
Checks on its own domain
| Check | Domain | Result | Scanned | Method |
|---|---|---|---|---|
| Ora cached score | snyk.io | 79 of 100, grade B | 5 November 2026 | how it was taken |
| Cloudflare agent-readiness scan | snyk.io | 40 of 100, L4 Agent-Integrated; 6 of 15 counted checks passed | 5 November 2026 | how it was taken |
What it does not do
- Agent Scan does not sit in the request path at runtime; it scans configurations and definitions.