Portkey

Open-source AI gateway and control plane for production Gen AI.

Hand-reviewed by the Toolglade team·Updated Aug 2026

coding#ai-gateway#llmops#open-source#observability

Free plan Claimed API Self-hosted Teams

Toolglade’s take

Portkey is a strong choice as a production gateway and control plane when you need to route across many models with governance, reliability, and observability. Its March 2026 decision to fully open-source the gateway, including governance features, is a meaningful differentiator for data control and avoiding lock-in. Keep in mind it adds a hop in the request path, log-volume pricing can grow at scale, and it is a gateway rather than a full build-and-eval IDE.

About Portkey

Portkey is an AI gateway and control plane for production generative AI that provides a single API to route across 1,600-plus models, plus observability, guardrails, governance, cost controls, key management, and prompt management. As of March 2026, its full gateway including governance is open source and self-hostable, while managed SaaS and Enterprise tiers remain. It operates at very large scale and suits engineering and platform teams running production workloads.

Portkey acts as a control plane for production generative AI. Through one unified API it routes requests across more than 1,600 models and providers, adds observability and tracing, enforces 40-plus guardrails, manages virtual keys and budgets, and handles reliability features like fallbacks, load balancing, caching, and retries. It also added an MCP Gateway to govern AI agents across enterprise tools. In March 2026, Portkey open-sourced its full gateway, including governance, observability, auth, and cost-control features that previously required the paid SaaS, plus the MCP Gateway. The gateway is self-hostable, while a managed SaaS and Enterprise tiers remain for hosted and advanced governance needs. The platform operates at very large scale, citing more than a trillion tokens and over 100 million requests per day across tens of thousands of organizations. Pricing is largely metered on log volume and retention, with a free developer tier, an affordable Production tier, and custom Enterprise.

TL;DR

Portkey is an AI gateway and control plane for production generative AI, offering one API to route across 1,600-plus models plus observability, guardrails, governance, cost controls, and prompt management. As of March 2026 its full gateway including governance is open source and self-hostable. It runs at very large scale and suits engineering and platform teams, with a free tier, a Production tier around $49 per month, and custom Enterprise.

Company overview

Portkey builds a production stack for generative AI centered on its gateway. It cites usage of more than a trillion tokens and over 100 million requests per day across tens of thousands of organizations, governing large amounts of annualized AI spend.

In March 2026 it open-sourced the full gateway, a notable move that put previously paid governance and observability features into the free, self-hostable product.

Product features

Portkey offers a unified API to 1,600-plus models across modalities, observability and tracing, more than 40 pre-built guardrails, prompt template management and a universal playground, routing with load balancing, fallbacks, and caching, virtual keys with budget and cost controls, and an MCP Gateway for agent governance.

The gateway is open source and self-hostable, while managed SaaS and Enterprise tiers add hosted convenience, SSO, VPC deployment, and data residency.

Target market

Engineering and platform teams and enterprises running production generative AI and agent workloads that need a governance, reliability, and observability control plane across many models, plus individual developers via the free tier and open-source gateway.

Buyer personas

End users

Backend and AI engineers integrating the gateway and configuring routing, guardrails, and observability.

Buyers

Platform engineering leaders and CTOs standardizing a control plane for production AI.

Key influencers

Security, governance, and FinOps stakeholders concerned with reliability, compliance, and cost.

Ideal customer profile

An organization running production generative AI across many models that needs governance, reliability, and observability in one gateway, with an open-source self-host option.

Funding & performance

Portkey has raised roughly $18M in total, including a $3M seed round in 2023 led by Lightspeed and a $15M Series A led by Elevation Capital with Lightspeed participating.

Pros & cons

Pros

Fully open-source, self-hostable gateway for data control and no lock-in
Access to more than 1,600 models through one API
Broad feature set spanning observability, guardrails, routing, prompts, and cost control
Transparent, relatively cheap entry with free and ~$49 tiers
Battle-tested at very large scale
Strong reliability tooling including fallbacks, load balancing, and caching
MCP Gateway addresses emerging agent-governance needs

Cons

Adds a network hop and dependency in the critical request path
Log-volume-based pricing can get complex or expensive at scale
Free-tier logging caps and short retention are tight for real debugging
Self-hosting the open-source gateway shifts ops burden to you
It is a gateway and ops layer, not a full build-and-eval IDE
Enterprise governance features are gated behind custom pricing
Relatively modest funding versus its scale and ambition

Pricing plans

Developer

~10,000 requests/mo
10,000 logs, 3-day retention
AI gateway and caching
Basic observability

Production

$49 / month

~100,000 requests/mo
~100,000 logs
Guardrails and load balancing
Team features

Enterprise

Custom

Unlimited logs
SSO
VPC or self-hosted deployment
Policy-as-code and data residency
Dedicated support

Key features

API

Team collaboration

Self-hosted

Integrations

1,600+ LLMs and providers, OpenAI, Anthropic, MCP Gateway, guardrails ecosystem

Input types

text, image

Output types

text, image

Best For

routing across many LLMs, production reliability and fallbacks, AI governance and guardrails, cost and usage control

Compare key features

View all alternatives →

Feature

Portkey

Dify

Helicone

Pricing

Freemium

Free plan

Yes

Free trial

API

Yes

Self-hosted

Yes

Team support

Yes

Frequently asked questions

Is Portkey open source?+

Yes. As of March 2026, Portkey open-sourced its full gateway, including governance, observability, auth, and cost-control features, plus the MCP Gateway. It is self-hostable from GitHub, while managed SaaS and Enterprise tiers remain for hosted needs.

How many models does Portkey support?+

Portkey provides a single unified API to route across more than 1,600 LLMs and providers across multiple modalities.

Does Portkey add latency?+

As a gateway it sits in the request path, which adds a network hop and a dependency. This is a trade-off for its routing, reliability, and observability features, and self-hosting can mitigate some concerns.

How does Portkey pricing work?+

Pricing is largely metered on log volume and retention. There is a free Developer tier, a Production tier around $49 per month, and custom Enterprise pricing. The open-source gateway is free to self-host.

What is the MCP Gateway?+

The MCP Gateway lets you govern AI agents as they access enterprise tools, adding policy, observability, and controls to agentic workflows.