What it does

Portkey is described on its site as a production stack for teams building with generative AI. Its gateway offers a unified API to more than 1,600 models, and the home page shows a short code sample using its Node.js SDK in place of calling a provider directly. Around the gateway sit observability dashboards for usage, cost and latency, guardrails, prompt management and a model catalog. For larger organizations, the site lists role-based access control, single sign-on, budget limits, activity and audit logs, and automatic redaction of sensitive data before requests reach a model. An MCP gateway centralizes authentication, access and observability for MCP servers. The site also says it has open-sourced the LLM pricing dataset it uses internally. The home page now announces that Portkey is the PRISMA AIRS AI Gateway, generally available for enterprises. Pricing is on a separate page that was not read, so it is not verified here.

Capabilities

Unified API to more than 1,600 LLMs
Observability for usage, cost and latency
Guardrails and automatic redaction of sensitive data
Prompt management and a model catalog
Access control, single sign-on, budgets and audit logs
MCP gateway for authentication and observability of MCP servers

When to use it

You want to call models from many providers through one API
You need to track and cap LLM usage and cost across teams
You must apply guardrails or redact sensitive data before requests reach a model
You need central access control over models and MCP servers