Depscope Mcp — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Depscope Mcp (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Package intelligence MCP server for AI agents. Stops AI coding agents (Claude, ChatGPT, Cursor, Windsurf, Copilot) from installing hallucinated, deprecated, or malicious packages across 19 ecosystems.
→ Backed by depscope.dev — 1.2M+ packages indexed, 19,000+ vulnerabilities tracked, real-time.
The MCP server now sends a system-prompt directive to your AI client at handshake (server.instructions). Claude Code, Cursor, Windsurf and other MCP clients receive a proactive-invocation brief automatically — manual rule files (CLAUDE.md, .cursorrules, .windsurfrules) are now optional. Existing rules still work; they're just redundant.
What the model sees at every session start:
check_malicious → check_typosquat → check_package → install_commandFor Claude Code there is also a companion plugin that bundles the MCP server with a skill carrying rich frontmatter triggers:
git clone https://github.com/cuttalo/depscope-claude-plugin ~/.claude/plugins/depscopeAll npm versions <0.9.0 are now deprecated. Run npm update -g depscope-mcp if you installed globally.
LLMs frequently invent package names that look real but don't exist (fastapi-turbo, lodahs, tokio-stream-extras). When an agent tries to install one, it might hit an attacker's typosquat. DepScope verifies every package before install.
Add to your MCP config:
{
"mcpServers": {
"depscope": {
"url": "https://mcp.depscope.dev/mcp"
}
}
}{
"mcpServers": {
"depscope": {
"command": "npx",
"args": ["-y", "depscope-mcp"]
}
}
}| Tool | Purpose |
|---|---|
check_package | Full package check: deprecated/CVE/health/recommendation |
get_health_score | 0-100 score with breakdown (maintenance/popularity/security/maturity/community) |
get_vulnerabilities | Open CVEs from OSV + KEV/EPSS |
package_exists | Hallucination detector (404 = LLM invented it) |
find_alternatives | Curated alternatives for deprecated/abandoned packages |
get_typosquat | Suspicious name similarity check |
get_breaking_changes | Migration plan between versions |
get_bugs | Known bugs from GitHub issues |
compare_packages | Side-by-side health/license/vuln comparison |
resolve_error | Map error message → likely cause + fix |
search_errors | Find similar error reports across ecosystems |
check_compat | Stack compatibility check |
get_latest_version | Latest stable + maturity signal |
| ... and 9 more | full list in tools.js |
npm · pypi · cargo · go · composer · maven · nuget · rubygems · pub · hex · swift · cocoapods · cpan · hackage · cran · conda · homebrew · jsr · julia
Free. No auth required. Generous rate limits. The MCP server is open-source (AGPL-3.0); the backend (depscope.dev API) is proprietary.
AGPL-3.0-or-later. Backend is proprietary; this client is open.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.