#ai

Compare the best open source and self-hosted AI gateways in 2026, including LiteLLM, Bifrost, Pangolin, Portkey Gateway, and Kong AI Gateway.

Learn what an AI gateway is, how it sits between applications, agents, and models, and how it differs from API, LLM, and MCP gateways.

Move your team off shared and per-user virtual API keys and onto identity-based access for your AI gateway - no keys to generate, distribute, or rotate for human users.

Virtual API keys solve model routing and budgeting well, but they were never built to be an identity system. Here's where that gap shows up in self-hosted LiteLLM and Bifrost deployments, and what to do instead.

A general-purpose guide to putting any self-hosted model server - vLLM, Ollama, llama.cpp, LM Studio, LocalAI, or a private downstream gateway - behind Pangolin's identity-aware AI gateway.

Run open-weight models like DeepSeek, Qwen, and Kimi K2 on your own hardware, then reach them securely from anywhere through Pangolin's AI gateway - no API keys, no data leaving your infrastructure.

Give engineers and coding agents access to a self-hosted vLLM or Ollama cluster from one endpoint, authenticated by identity instead of API keys passed around a team.

Pangolin 1.22 introduces an AI Gateway for cloud and self-hosted models with identity-aware access. SSH, RDP, and VNC public resources, plus SSH and HTTPS private resources, are now available in Community Edition.

Access your NVIDIA DGX Spark from any laptop without SSH, a VPN, or an open port. Run Ollama on the Spark and reach it through Pangolin's identity-aware AI gateway, from the same endpoint you use for cloud models.

Learn what MCP tunnels are, how they connect AI agents to private Model Context Protocol servers over outbound-only connections, and why they matter for enterprise security.