Portkey AI Gateway
Open-source gateway routing requests to 250+ LLMs through one API, with retries, fallbacks and load balancing.
MIT-licensed router with fallbacks, load balancing and guardrails you run yourself.
Portkey AI Gateway is a lightweight router that sends requests to many language model providers through one API. It suits developers who want fallbacks and load balancing without a hosted middleman.
What it is
The gateway routes to 250+ LLMs (the README also cites 1600+ models) with automatic retries, fallbacks, load balancing, conditional routing and guardrails. It runs locally with npx @portkey-ai/gateway, and the README says a 2.0 pre-release is merging Portkey's enterprise gateway into open source.
Who it's for
It suits developers and platform teams who want to control routing between providers. It does not suit anyone wanting dashboards and spend reports from the open-source repo alone, since Portkey sells its hosted and enterprise products separately.
Key features
250+ LLM routing
Reach many providers through one API.
Retries and fallbacks
Automatic retries and fallback to another model on failure.
Load balancing
Spread requests across keys or models with conditional routing.
Guardrails
Apply checks to requests and responses.
MCP gateway docs
Portkey documents MCP server management with auth and observability.
Small footprint
The README quotes a 122kb install size.
Up and running in 3 steps
Our honest verdict
What's great
- One npx command starts a local gateway.
- MIT licence covers use in commercial services.
- Fallbacks and load balancing are built in.
- The 2.0 pre-release branch is public for early access to enterprise features.
Where it falls short
- The last commit on the default branch was in May 2026, slower than most gateways.
- Some enterprise features sit in Portkey's paid products until 2.0 ships.
- You operate and secure the server yourself.
Pricing & license
The gateway is free under MIT. Portkey also sells a hosted gateway and enterprise plans on portkey.ai; this repo carries no price.
Open remote AI-training roles
870 open remote AI-training roles, updated November 14, 2026; most pay $20–$170 an hour. Zekai lists them free; the application goes to the employer’s platform.
Alternatives
These reviewed AI tools do the same job instantly — from our scored directory.
Use it, or have us set it up
FAQ
Is it the same as Portkey the hosted service?
No. The repo is the open-source gateway; Portkey also sells hosted and enterprise versions.
How do I run it?
Use npx @portkey-ai/gateway with Node.js and npm, or Docker.
What is 2.0?
A pre-release branch that merges Portkey's core enterprise gateway into the open-source project, per the README.
