Back to All Tools
⚡ Model Context Protocol (MCP) ServerMCP Servers16 StarsVerified Source (98%)
Owner Email Not Available
No Public Author Email FoundNo public contact email was found in Git commit logs or GitHub profile. To link your email, configure your public Git commit email or GitHub profile and trigger a Re-sync from GitHub.

Agentshield

🛡️ The security firewall for AI agent tools & MCP servers (Claude, GPT-5.6, Gemini 3.7, Antigravity, Cursor, Codex, Hermes). Catch command injection, secret theft & toxic flows in <50ms. 100% offline Rust SAST + 1-click auto-fix + SARIF.

#cloudflare-os#mcp-servers
Verified Client Compatibility:
Claude DesktopCursor IDEWindsurfCline / RooZed EditorContinue.dev
Deep Inspect in MCP Inspector
Visit Official Website
Synced from GitHub HomepageWebsite URL is fetched automatically from the GitHub repository "About / Website" field. We provide do-follow backlink for this URL. To update this link, edit your repository settings on GitHub and click "Re-sync from GitHub".
View on GitHubDeploy to Cloudflare

AST Security Audit & Verification Scorecard98% Score

Last verified & scanned on: Aug 20, 2026

Manifest CheckPython
License AuditPermissive Software License
Supply-Chain Risk0 High-Risk Mutation CVEs

5-Pillar Verification Breakdown (100-Point Standard)

Total Score: 98 / 100 pts
Edge Sandbox Safety
18 / 20 pts

Sandboxed edge isolate runtime safety verified.

Ecosystem & MCP Alignment
20 / 20 pts

Model Context Protocol tool alignment verified.

Open-Source License Compliance
20 / 20 pts

License verified as Other.

Documentation & Quickstart Quality
20 / 20 pts

Documentation and usage guides inspected.

Repository Hygiene & Provenance
20 / 20 pts

Active Git provenance verified.

How to Improve This Score to 100/100

To increase this repository's score from 98% to 100%, author and contributors can apply these updates to GitHub:

  • +2 pts:Edge Sandbox SafetyEnsure code avoids dynamic eval(), unsanitized child_process spawns, and hardcoded secrets.
Once updates are pushed to GitHub, trigger a deep scan above after the cooldown period.View Full Scoring Rubric →

Universal Client Configurations & Quickstart

Select your environment below for instant 1-click CLI installer or verified JSON client configs

Select Installation Method or AI Client:
1-Click Terminal Command
# 1-Click install into Claude Desktop (auto-injects into claude_desktop_config.json):
npx unofficialos add agentshield

# 1-Click install into Cursor IDE (.cursor/mcp.json):
npx unofficialos add agentshield --client cursor

# Or run directly via terminal:
npx -y github:aiconnai/agentshield