MCP Gateway
A plugin-based MCP gateway that sits between an LLM and other MCP servers, intercepting and sanitizing requests/responses to prevent PII leakage, token/secret exposure, prompt injection attacks, and harmful content — with built-in security scanning of MCP server reputation.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
MCP API gateway/proxy. Auth and rate limiting at gateway layer. Gateway sees all tool traffic — high trust requirement. TLS termination at gateway. Audit logs essential.
⚡ Reliability
Best When
You are deploying MCP servers in an enterprise context and need a drop-in security layer for PII masking, secret detection, and prompt injection defense without rewriting existing MCP servers.
Avoid When
You want full security coverage (PII + secrets + prompt injection + harmful content) without the Lasso commercial plugin — the open-source plugins only partially cover these cases.
Use Cases
- • Adding a security and PII protection layer in front of existing MCP servers without modifying them
- • Detecting and blocking prompt injection attempts embedded in MCP tool responses before they reach the LLM
- • Auditing and logging MCP tool calls for compliance and observability via the xetrack plugin
Not For
- • Simple personal MCP setups where security overhead is unnecessary
- • Teams that need full enterprise policy management without Lasso Security's commercial API key (basic/presidio plugins have limited coverage)
- • High-throughput production environments where the Python proxy layer adds latency concerns
Interface
Authentication
Lasso plugin requires LASSO_API_KEY. The gateway itself has no auth — it relies on the underlying MCP client's auth. xetrack uses path-based config.
Pricing
MIT licensed core; Lasso plugin requires commercial API key for full security coverage.
Agent Metadata
Known Gotchas
- ⚠ Silent redaction of PII/secrets means agents may receive incomplete tool responses without understanding why — the agent may not realize data was removed
- ⚠ Full security coverage (prompt injection + harmful content detection) requires the commercial Lasso Security API plugin — open-source plugins only cover PII via Presidio and basic secret patterns
- ⚠ Security reputation scanner checks MCP servers at gateway load time but does not continuously re-scan — a server compromised after initial startup will not be detected
- ⚠ Python proxy adds a latency hop for every MCP call — not benchmarked; may be significant for high-frequency tool calls
- ⚠ The gateway itself has no authentication layer — it relies on the MCP client to be the trust boundary, which may not be appropriate for multi-tenant environments
Alternatives
Full Evaluation Report
Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for MCP Gateway.
Scores are editorial opinions as of 2026-03-06.