Portkey Open-Sources Its AI Gateway After Processing 2 Trillion Tokens a Day

Why It Matters

By open‑sourcing the core gateway, enterprises can avoid costly SaaS lock‑ins while gaining enterprise‑grade observability and policy enforcement for large‑scale AI and agentic workloads, accelerating trustworthy AI adoption.

Key Takeaways

• Open-source gateway manages 2 trillion tokens daily.
• Handles 120 million AI requests across 24,000 organizations.
• Provides built‑in governance, cost controls, and MCP support.
• Aims to scale token processing 1,000× by 2027.
• Reduces reliance on multiple SaaS contracts for AI infrastructure.

Pulse Analysis

The rise of AI gateways reflects a broader shift toward modular, observable infrastructure that can keep pace with the massive token volumes generated by production models. Traditional API gateways lack the nuanced policy enforcement required for large language models and autonomous agents, prompting vendors to build dedicated control planes. Open‑source offerings like Portkey’s Gateway lower the barrier to entry for engineering teams, allowing them to integrate governance, authentication, and cost‑tracking directly into their CI/CD pipelines without negotiating separate SaaS contracts.

Portkey’s operational metrics underscore the urgency of such solutions. Processing two trillion tokens daily and supporting over 120 million requests, the platform already manages roughly $180 million in annual AI spend for 24 k enterprises. Its new features—real‑time cost and latency dashboards, OAuth 2.1 authentication, and a model catalog spanning multiple providers—provide the visibility and control that large organizations need to prevent budget overruns and compliance breaches. The inclusion of an MCP gateway further differentiates the product, giving teams a secure, observable layer for agentic workflows that can act on internal systems.

For the market, Portkey’s open‑source strategy could catalyze a wave of community‑driven enhancements, driving down the total cost of ownership for AI production stacks. Competitors may feel pressure to either open their core components or bundle more value‑added services, reshaping the SaaS landscape. If the company achieves its 1,000‑fold token growth target by 2027, the gateway could become the de‑facto reference architecture, compelling cloud providers and AI platform vendors to integrate compatible control‑plane APIs, thereby standardizing governance across the AI ecosystem.