Last updated 06-27-2026
Category:
Reviews:
Join thousands of AI enthusiasts in the World of AI!
Portkey
Portkey is a production stack for teams shipping generative AI applications. It combines an AI gateway, observability, guardrails, governance, and prompt management in one platform so developers can route LLM traffic, monitor costs, and enforce safety policies without stitching together separate tools.
The platform centers on the AI Gateway pattern: a single unified API that connects to 1600+ models and providers across text, vision, audio, and image generation. Smart routing handles fallbacks, load balancing, retries, and timeouts, while virtual keys keep provider credentials in a managed vault instead of scattered across codebases.
Portkey is built for ML engineers, platform teams, and enterprises running GenAI in production. Customer stories cite use cases from cost attribution across dozens of LLM apps to caching that cuts repeated test runs in CI pipelines. The company is open source, HIPAA compliant, and powers more than 3,000 GenAI teams.
One unified API routes traffic to 1600+ LLMs and multimodal providers
Automatic fallbacks, load balancing, retries, and request timeouts built in
Logs every request with 40+ details on cost, latency, and accuracy
60+ guardrails block prompt injections and redact PII in real time
Prompt studio with versioning, variables, playground, and API deployment
Simple and semantic caching to cut repeat inference spend
Open-source gateway you can self-host with community support
Unified API covers 1600+ models so teams avoid per-provider integration work.
Free tier and open-source gateway make it easy to prototype before committing.
Observability logs 40+ metrics per request for cost attribution and debugging.
Enterprise plan includes HIPAA, SOC2, GDPR compliance and private cloud deployment.
Free plan caps at 10k recorded logs per month and is not meant for production.
Production plan lacks custom security controls and data residency guarantees.
Enterprise pricing requires a sales conversation with no public rate card.
Does Portkey have a free plan?
Yes. Portkey offers a Free Forever plan with 10,000 recorded logs per month, 3-day log retention, AI gateway routing, observability, 3 prompt templates, simple caching, and deterministic guardrails. It is intended for prototyping and evaluation, not production workloads.
How much does Portkey cost for production use?
Portkey's Production plan costs $49 per month and includes 100,000 recorded logs per month, with $9 overages per additional 100,000 requests. It adds LLM and partner guardrails, unlimited prompt templates, role-based access control, semantic caching, and production support.
How many LLM providers does Portkey support?
Portkey connects to 1600+ LLMs and providers across different modalities through its unified AI gateway API. The homepage also cites support for 148+ LLMs in its stats section, with new models added as providers release them.
Can Portkey be self-hosted?
Yes. Portkey offers an open-source AI gateway you can deploy locally. The pricing page lists a Host it Yourself option with universal API, retries, routing, guardrails, fallbacks, load balancing, and community support.
Does Portkey add latency to LLM API requests?
Portkey markets its AI gateway as adding 0% latency overhead for real-time AI experiences. Its feature pages emphasize lightning-fast responses and battle-tested infrastructure designed for production reliability.
Does Portkey support SSO and team management?
Yes. Portkey's Enterprise plan includes SSO with Okta Auth, role-based access control, team management, granular budget and rate limits, and org-wide audit logs. The Production plan also includes RBAC and service account API keys.
What guardrails does Portkey offer?
Portkey runs 60+ guardrails on top of its open-source gateway, including input checks for prompt injection and output filters for PII leaks and hallucinations. It integrates partners like Mistral, Patronus, Pangea, Bedrock, Azure, and Palo Alto Networks.