Portkey
Portkey equips AI teams with an AI Gateway, observability, guardrails, and prompt management to productionize generative AI applications.
Last verified:
What is Portkey?
Portkey is a comprehensive AI control panel and gateway platform designed to help GenAI teams build, deploy, and manage production-ready AI applications. It provides a unified API that connects to over 1,600 LLMs from 200+ providers, eliminating the need to manage multiple API keys and provider-specific integrations. The platform sits between your application and LLM providers as a forward proxy, handling routing, fallbacks, caching, guardrails, observability, prompt management, and AI governance through a single interface.
Key features include the AI Gateway with intelligent load balancing, automatic retries, conditional routing, and semantic caching; real-time observability with 50+ metrics per request including latency, cost, token usage, and hallucination risk; 50+ AI guardrails for safety, PII redaction, prompt integrity, and output quality enforcement; collaborative prompt management with versioning, testing, and optimization across models; and enterprise AI governance with RBAC, SSO, budget limits, rate limits, and detailed activity logs. The platform integrates in just 3 lines of code without changes to your existing stack.
Portkey is designed for AI developers, machine learning engineers, AI teams at startups and Fortune 500 companies, and enterprises building production-grade GenAI applications. It powers over 3,000 GenAI teams and processes 2.5 trillion tokens daily, with 99.99% uptime serving over 25 million requests daily. The platform is HIPAA compliant, SOC 2 Type II certified, ISO 27001 certified, and GDPR compliant, making it suitable for regulated industries.
Portkey pricing
Pricing model: Freemium
Free tier: 10,000 recorded logs per month with 3-day log retention and 30-day metric retention, includes AI Gateway (Universal API, fallbacks, load balancing, retries), basic observability (logs, traces, feedback, custom metadata, filters), 3 prompt templates with playground and API endpoints, simple caching, and deterministic guardrails. Developer plan includes community support. Production plan: 100,000 recorded logs per month at $59.99/month, plus $9 overage per additional 100k requests, with 30-day log retention and 90-day metric retention, includes unlimited prompt templates, LLM and partner guardrails, alerts, semantic caching, RBAC, and service account API keys. Enterprise plan: 10 million+ recorded logs per month with custom pricing, custom retention periods, custom guardrail hooks, advanced evaluation templates, SSO, granular budget and rate limits, private cloud deployment, VPC hosting, advanced compliance (SOC2 Type 2, GDPR, HIPAA), custom BAAs, data isolation, and dedicated onboarding with priority support. Open-source AI Gateway is free to self-host.
Portkey pros
- Unified API access to 1,600+ LLMs from 200+ providers
- Integrates in just 3 lines of code with no stack changes
- Intelligent load balancing across multiple models
- Automatic fallbacks when providers fail
- Semantic caching reduces costs and latency significantly
- 50+ real-time AI guardrails for safety and compliance
- Automatic PII redaction before sending to LLM
- Real-time observability with 50+ metrics per request
- Granular cost tracking down to individual user level
- Collaborative prompt management with versioning
- Conditional routing based on cost, latency, or specialization
- Role-Based Access Control for team management
- Single Sign-On integration with custom OIDC providers
- 99.99% uptime serving 25M+ requests daily
- HIPAA, SOC 2 Type II, ISO 27001, and GDPR compliant
- Handles millions of requests per minute with high concurrency
- Smart retry logic triggers automatically on LLM failures
- Budget limits and rate limits per team/environment
- Open-source AI Gateway available for self-hosting
- Edge workers globally distributed for minimal latency addition
Portkey cons
- Adds 20-40ms latency per request compared to direct API calls
- Usage-based pricing at high volumes can be unpredictable
- Less flexible than building fully custom DIY stack
- Some advanced features require paid plans only
- Semantic caching accuracy depends on query similarity thresholds
- Enterprise features like private cloud deployment cost extra
- Gateway benchmark tests show higher latency than optimized alternatives like Kong
- Limited customization for highly specialized routing logic
- May be overkill for simple single-model applications
- Guardrail configuration requires some learning curve
Frequently asked questions about Portkey
What is Portkey?
Portkey is a comprehensive AI control panel and gateway platform that serves as a unified interface for interacting with over 1,600 AI models from 200+ providers. It provides AI Gateway, observability, guardrails, governance, and prompt management all in one platform. Portkey sits between your application and LLMs as a forward proxy, handling routing, fallbacks, caching, guardrails, and monitoring to make your AI apps resilient, secure, performant, and more accurate.
Will Portkey increase the latency of my API requests?
Portkey is hosted on edge workers throughout the world, ensuring minimal latency. Their benchmarks estimate a total latency addition between 20-40ms compared to direct API calls. This slight increase is often offset by the benefits of their caching and routing optimizations. In optimized production environments, latencies can be as low as 1-2ms.
Is my data secure?
Portkey AI is ISO:27001 and SOC 2 certified, and GDPR & HIPAA compliant. All data is encrypted in transit and at rest using industry-standard AES-256 encryption. For enhanced security, they offer a feature that does NOT store request and response body objects in Portkey datastores or logs. For enterprises, they offer managed hosting to deploy Portkey inside private clouds.
Will Portkey scale if my app explodes?
Portkey is built on scalable infrastructure and can handle millions of requests per minute with very high concurrency. They currently serve over 25M requests daily with a 99.99% uptime. Their edge architecture and scaling capabilities ensure they can accommodate sudden spikes in traffic without performance degradation.
Does Portkey impose timeouts on requests?
Portkey does NOT impose any explicit timeout for free OR paid plans currently. While they don't time out requests on their end, they recommend implementing client-side timeouts appropriate for your use case to handle potential network issues or upstream API delays.
Do you support SSO?
Yes! Portkey supports SSO with any custom OIDC provider. This is available on Enterprise plans and allows teams to onboard instantly while following governance rules from day 0.
What are the pricing options? Is there a free trial?
Portkey's Gateway is open source and free to use for self-hosting. On the managed version, Portkey offers a free plan with 10k requests per month. They also offer paid Production plan at $59.99/month for 100k logs and Enterprise plan with custom pricing for 10M+ logs. All plans include the core AI Gateway features.
Can we prevent storage of sensitive data?
Yes. On request, Portkey can enable a feature that does NOT store any of your request and response body objects in Portkey datastores or logs. Additionally, Portkey automatically redacts sensitive PII data from requests before they are sent to the LLM using their PII redaction guardrail.
How many providers do you support?
Portkey supports over 200 providers and provides access to 1,600+ LLMs through their unified API. This includes OpenAI, Anthropic, Groq, Hugging Face, Microsoft Azure, MongoDB, and many more major providers without requiring code changes.
Can we self-host Portkey?
Yes. Portkey's AI Gateway is open source and can be deployed locally or in your own infrastructure. Self-hosting includes Universal API, retries & timeouts, routing, guardrails, automatic fallbacks, basic dashboard, load balancing, and community support. This is free to use, though you bear the infrastructure and maintenance costs.