Skip to content
Stork.AIstork.ai
AI ToolBecomes the API

Portkey AI Gateway Review

Portkey AI Gateway is a switchboard for Large Language Model (LLM) endpoints that handles key management, retries, caching, and AB routing.

shipped Nov 20, 2025buildpaid
BuildServingInference Gateways
Portkey AI Gateway — product screenshot

Why it matters

1Supports over 1,600 models from more than 40 providers, including OpenAI, Anthropic, and Google.
2Acquired by Palo Alto Networks in April 2026 to become the AI Gateway for Prisma AIRS.
3The unified Gateway, including governance and observability features, was made open source in March 2026.
4Offers a free tier for initial usage and provides a developer API.

Stork’s verdict on Portkey AI Gateway

Portkey is a solid LLM gateway for production, but its dashboard can be slow under heavy loads.

Portkey AI Gateway reviewed by Stork AI · stork.ai/en/portkey-ai-gateway

Stork Quadrant

Becomes the API· 35/100

Replaceable as a UI, but kept alive as the API the agents call.

“Portkey is infrastructure glue, not a moat. Every capability it offers is a weekend project for a competent engineer, and the major cloud providers are building this natively into their AI SDKs. The coordination moat is thin — it connects N providers and N apps, but there's no liquidity lock-in like a real marketplace. This category gets commoditized fast.”

— Claude Sonnet 4.6, scored 2026-05-30

Defensibility · 15/100

  • Physical-world coupling
  • Regulatory moat
  • Network liquidity
  • Proprietary refreshing data
  • High-trust catastrophic workflows
  • Multi-party coordination
  • Brand / community / taste

An LLM alone could replace

  • Route a request to a fallback LLM provider if the primary fails — any retry logic can be hand-coded in an afternoon
  • Cache identical prompts to avoid redundant API calls — a Redis layer does this without Portkey
  • Manage and rotate API keys across providers — a secrets manager like Vault or AWS Secrets Manager handles this
  • A/B test two model versions by splitting traffic — a feature flag system already does this

Agent-Readiness · 60/100

  • Verified MCP
  • Listed on agent surfaces— anthropic_directory
  • Usage-based pricing
  • Headless agent auth— https://portkey.ai/docs/introduction/what-is-portkey?utm_source=website&utm…
  • Public OpenAPI— https://portkey.ai/docs/introduction/what-is-portkey?utm_source=website&utm…
  • Active changelog— https://portkey.ai/blog/enterprise-need-for-an-llm-gateway/ (2026-04-20)
  • llms.txt— https://www.portkey.ai/llms.txt

Score history · +13 pts over 2 re-scores

How to defend

Stop being a dashboard and become the observability layer that enterprises can't rip out — own the audit logs, cost attribution, and compliance reporting that procurement teams require. Alternatively, go deep on a regulated vertical where LLM routing decisions need to be logged and explainable for legal reasons.

  • Ship an MCP server and list it on Stork — biggest single point gain (+25).
  • Add a usage-based or per-call tier; per-seat-only pricing dies when agents replace seats (+15).
How this score is computed →See the full quadrantHow to defend

Specs

API Docs

View Documentation →

GitHub

View Repository →

API Available

Yes, public API

overview

What is Portkey AI Gateway?

Portkey AI Gateway is an AI Gateway tool developed by Portkey that enables Gen AI Builders and AI teams to simplify the integration, management, and scaling of AI models in production environments. It acts as a control plane between applications and various model providers, offering a unified interface and production-ready features. The platform provides a single unified endpoint to access over 1,600 models from more than 40 providers, including OpenAI, Anthropic, Google, and Azure OpenAI. It is designed to move AI applications from experimentation to stable production by offering visibility, control, and governance over LLM usage, ensuring high availability and optimal performance through features like automatic retries, fallbacks, load balancing, and caching. A significant development for Portkey AI Gateway is its acquisition by Palo Alto Networks, announced in April 2026, with plans for Portkey to become the AI Gateway for Prisma AIRS.

features

Key Features of Portkey AI Gateway

Portkey AI Gateway provides a comprehensive suite of features designed for robust LLM orchestration in production. These capabilities streamline model integration, enhance reliability, and provide granular control over AI application performance and costs.

  • Unified API access to over 1,600 models from 40+ providers.
  • Automatic key management and rotation for LLM API keys.
  • Intelligent retries and fallbacks to ensure application resilience and high availability.
  • Semantic caching for repeat or near-identical queries, reducing latency and API costs.
  • A/B routing capabilities for comparing and testing different LLM models or prompts.
  • Comprehensive observability with centralized logs, request tracking, cost insights, and performance metrics.
  • Guardrails and governance features, including PII redaction and Role-Based Access Control (RBAC).
  • Prompt management and versioning for consistent and controlled prompt deployment.
  • Model Context Protocol (MCP) Gateway for managing and governing AI agentic workflows.
  • Hybrid deployment options, allowing the gateway to run within a user's network while the control plane remains SaaS.

use cases

Who Should Use Portkey AI Gateway?

Portkey AI Gateway is designed for various stakeholders involved in developing and deploying AI-powered applications, particularly those requiring robust, scalable, and observable LLM infrastructure.

  • Gen AI Builders and developers integrating diverse LLM providers into their applications.
  • AI teams scaling applications from experimentation to stable production environments.
  • Organizations needing to manage, observe, and govern GenAI usage across multiple internal teams.
  • Enterprises requiring enhanced security and compliance, including PII redaction and data processing addendums.
  • Developers building with the Model Context Protocol (MCP) for advanced AI agent governance and tool orchestration.

how to use

How to Use Portkey AI Gateway

Getting started with Portkey AI Gateway involves integrating its unified API into your application and configuring your LLM providers through its control plane. The platform is designed for quick setup, often advertised as requiring only a few lines of code.

  • 1Sign up for a Portkey account via their website (www.portkey.ai).
  • 2Integrate the Portkey API into your application, replacing direct calls to individual LLM providers.
  • 3Configure your desired LLM providers (e.g., OpenAI, Anthropic) and their respective API keys within the Portkey dashboard.
  • 4Utilize Portkey's features such as caching, retries, and A/B routing by configuring them through the control plane.
  • 5Monitor LLM usage, costs, and performance in real-time using the centralized observability dashboard.
  • 6Implement guardrails, PII redaction, and governance policies to manage and secure production AI applications.

pricing

Portkey AI Gateway Pricing & Plans

Portkey AI Gateway operates on a paid model, with a free tier available for initial usage. The vendor website advertises a free tier, allowing users to explore core functionalities before committing to a paid plan. In March 2026, Portkey made its unified Gateway completely open source, including features previously requiring a SaaS subscription such as governance, observability, authentication, and cost controls. Specific pricing figures for paid tiers are not publicly itemized on the website, requiring direct inquiry for detailed enterprise or high-volume plans. The pricing page (https://www.portkey.ai/pricing) provides general information on available plans.

  • Free Tier: Available with certain usage limits.
  • Paid Plans: Specific details and pricing require direct inquiry.

Pros

  • +Provides a unified API for over 1,600 models from 40+ providers, simplifying multi-LLM integration.
  • +Offers comprehensive observability with centralized logs, request tracking, cost insights, and performance metrics.
  • +Ensures high reliability and resilience through automatic retries, fallbacks, and load balancing.
  • +Facilitates significant cost optimization via semantic caching and intelligent routing.
  • +Includes robust governance and guardrail capabilities, such as PII redaction and Role-Based Access Control (RBAC).
  • +The unified Gateway became open source in March 2026, enhancing transparency and control for users.

Cons

  • −The dashboard can experience slowness when processing heavy trace loads.
  • −Documentation for advanced routing features could be more comprehensive for complex configurations.
  • −Free tier limits may be quickly reached with real-world usage volumes, necessitating an upgrade.
  • −The extensive feature set might be overwhelming for newcomers or those seeking a simpler solution.
  • −Specific pricing details for paid tiers are not publicly itemized, requiring direct inquiry.
  • −The acquisition by Palo Alto Networks in April 2026 may lead to shifts in product roadmap or independent development.

Policies

Pricing Page

View Pricing→

Similar Tools

Portkey AI Gateway vs Competitors

Portkey AI Gateway positions itself as an end-to-end LLM orchestration platform, offering a 'Control Panel for Production AI.' It differentiates itself by providing a comprehensive suite of features beyond simple routing, with a strong emphasis on observability, governance, and guardrails for production environments.

1
LiteLLMOn StorkCompare

LiteLLM provides an OpenAI-compatible API for over 100 LLM providers, allowing seamless switching between models without code changes.

Compared to Portkey, LiteLLM is primarily an open-source, self-hosted solution, offering extensive provider coverage and built-in cost tracking, budgets, and fallbacks. While Portkey offers a managed platform with strong production guardrails, LiteLLM is ideal for teams prioritizing self-hosting and maximum flexibility.

2
TrueFoundryOn StorkCompare

TrueFoundry is a full-stack AI infrastructure platform that includes a high-performance LLM Gateway, offering comprehensive MLOps capabilities beyond just routing.

TrueFoundry's LLM Gateway provides similar routing, caching, and observability features to Portkey, but it extends to full deployment control, allowing teams to self-host open-source models and manage the entire GenAI application lifecycle. Portkey focuses more on middleware gateway functionalities for external APIs, whereas TrueFoundry offers a broader platform for building, deploying, and managing AI applications.

3
Cloudflare AI GatewayOn StorkCompare

Cloudflare AI Gateway offers a zero-infrastructure, managed solution leveraging Cloudflare's global edge network for caching and performance.

Cloudflare AI Gateway provides core gateway features like caching, rate limiting, and analytics with a generous free tier, making it suitable for teams already within the Cloudflare ecosystem. While Portkey offers deeper observability and prompt management, Cloudflare's solution excels in ease of setup and edge performance without requiring additional infrastructure.

4
HeliconeOn StorkCompare

Helicone is an observability-first, open-source AI Gateway that provides detailed request logging, prompt management, and cost attribution.

Helicone's primary strength lies in its robust observability features, offering in-depth tracing, session analytics, and cost attribution purpose-built for LLM workloads. While Portkey also offers strong observability, Helicone is often highlighted for its open-source nature and tight integration of monitoring into the gateway layer, making it a strong alternative for teams prioritizing analytics and debugging.

Visit Portkey AI Gateway↗
Connect
𝕏
X / Twittertwitter.com/portkeyai
⌘
GitHubgithub.com/portkey-ai
in
LinkedInlinkedin.com/company/portkey-ai