Portkey
App : A product you use directly
AI gateway with routing, guardrails, and observability, open source since March 2026.
The best governance/simplicity balance for a product team.
Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.
What it does, concretely
Teams using multiple language models face a concrete problem: each provider has its own API, its own keys, and its own response formats. Portkey solves this by placing a single gateway in front of all these providers, accessible through an OpenAI-compatible API.
In practice, Portkey intercepts each request and applies rules defined in a configuration: automatic retries on failure, fallback to another model if the first is unavailable, load balancing across multiple API keys, and guardrails that filter inputs and outputs according to customizable criteria. The latency added by the gateway is under 1 millisecond. Over 1,600 text, vision, audio, and image models are accessible through this unified interface.
Portkey is designed for product teams and developers deploying AI applications in production who need reliability, traceability, and cost control. The platform is used by more than 3,000 teams and processes billions of tokens every day.
However, Portkey is not suited for personal or occasional use: getting started requires development knowledge, and advanced features such as collaborative prompt management, semantic caching, or automatic provider optimization are reserved for paid plans or the hosted version.
What it offers
- ·Unified routing to over 1,600 language, vision, audio, and image models through a single OpenAI-compatible API.
- ·Automatic retries on failure, with up to 5 attempts and an exponential backoff strategy to prevent network overload.
- ·Automatic fallback to another provider or model on error, with the ability to target specific error codes.
- ·Load balancing across multiple API keys or providers with configurable weights.
- ·Input and output guardrails: over 40 pre-built checks to filter non-compliant content.
- ·Real-time observability: dashboard tracking requests, latency, costs, and error rates.
- ·Collaborative prompt management with versioning and a shared testing playground (paid).
- ·Semantic caching to reduce costs by avoiding redundant model calls (paid).
In practice
- Price
- Freemium
- Commercial use
- Commercial use: yes
- Available on
- web, docker
- Licence
- Apache-2.0
Security and data
- Rating
- Security: nothing to flag
- Your data
- Local or cloud depending on use
- Maintenance
- Active
- Last release
- 12 January 2026
Links and repository checked on 25 October 2026 · added on 1 September 2023
How to install Portkey
npx @portkey-ai/gateway