MCP Firewall
Prompt injection firewall middleware that sits between client and server as a stdio proxy, inspecting JSON-RPC requests for injection patterns and blocking suspicious requests.
About
Provides prompt injection detection middleware for MCP servers by acting as a stdio proxy between client and server. Inspects every JSON-RPC request against 12 built-in detection patterns across 6 attack categories including classic prompt injection, role hijacking, encoded payloads, and exfiltration attempts. Features a configurable scoring engine with pass/warn/block thresholds, YAML-based configuration with custom patterns, allowlists, and dry-run mode for testing.
Is this your project?
Claim this listing to manage your page, access analytics, and unlock upgrades. Verification takes 60 seconds.
Share This Project
Embed Badge
Add this badge to your README:
[](https://hifriendbot.com/ai-list/mcp-firewall/)
