MCP Gateway is an extension of the AI Gateway platform that secures Model Context Protocol (MCP) traffic between AI agents and MCP servers.
MCP Gateway runs inline between MCP clients and MCP servers (tools and data providers). It routes all MCP traffic through the AI Gateway appliance so you can apply the same controls you use for LLM traffic: access control, DLP, AI Guardrails, rate limiting, and audit logging.
Avantages
MCP Gateway extends AI Gateway controls to the broader agentic ecosystem.
- MCP server registration: Register MCP servers by name, URL, and protocol in the WebUI or by using the REST API.
- Access control policies: Allow or block traffic by server, tool, resource, prompt, activity type, or token group.
- DLP inspection: Inspect tool call arguments, tool call results, and resource content before data reaches the wrong destination.
- AI Guardrails: Evaluate tool inputs and outputs against your AI safety and compliance rules.
- Rate limiting: Apply per-server and per-token-group limits separately from LLM provider rate limits.
- OAuth2 transparent relay: Relay OAuth2 discovery metadata while credentials pass through without being stored.
- SkopeIT events: Examiner les champs spécifiques au MCP dans la table d'événements AI Gateway .
- SSE streaming inspection: Process server-sent event streams frame by frame. On by default for MCP traffic.