Portkey was founded in 2023 by Rohit Agarwal and Ayush Garg to give engineering teams a single, production-grade control point for everything that happens between their applications and the growing universe of language models. As companies began calling many providers and models, they ran into fragmented APIs, unpredictable latency and failures, runaway costs, and almost no visibility or governance. Portkey was built to consolidate all of that into one gateway and control plane.
At its core is a high-performance AI gateway that routes requests to more than 1,600 models through a single, OpenAI-compatible interface. The gateway provides reliability primitives that matter in production: automatic retries, fallbacks across providers, load balancing, request timeouts, and semantic caching to cut cost and latency. On top of routing, Portkey layers observability with detailed logs, traces, and analytics for every request, so teams can see token usage, costs, latencies, and error rates across models and applications.
The platform also addresses governance and safety. It offers budget limits, rate limits, virtual API keys, role-based access controls, and integrated guardrails that can validate inputs and outputs against policies. This combination of routing, observability, reliability, and governance is what Portkey calls the control plane for production AI, and it is open-source at the gateway layer with a managed cloud and enterprise offering on top.
Portkey raised a $3 million seed round led by Lightspeed in 2023, followed by a $15 million Series A led by Elevation Capital with continued participation from Lightspeed, bringing total funding to roughly $18 million. The company reports processing well over 500 billion tokens across more than 100 million requests per day for over 24,000 organizations, including AI-forward enterprises such as Postman and Snorkel AI.
Portkey competes with other AI gateways and LLMOps platforms but stands out by combining a fast open-source gateway with enterprise governance and full-stack observability, making it a strong fit for teams that need to operate many models reliably and within policy at scale.