Portkey AI

What Portkey AI is

Portkey AI is a platform for building and running generative AI applications. It provides a unified AI gateway across a large number of models, plus observability, prompt management and an MCP client, with load balancing, conditional routing, automatic retries and semantic caching.

It offers real-time monitoring and log tracking, and supports open source architecture with on-premises deployment.

What you can do with it

• Call many models through one gateway
• Add routing, caching and retry behaviour
• Track cost, latency and errors
• Self-host for data control

Who it is for

• Teams building generative AI products
• Organisations unifying several model vendors
• Enterprises with security and compliance needs

What to watch out for

• Prompts and responses pass through the gateway, so its logging settings decide what is stored
• The model count on the site is a vendor figure; confirm the providers you need are supported
• Semantic caching can return stale or wrong answers if thresholds are loose
• Self-hosting adds operations work