Portkey Router: AI Gateway and Multi-LLM Control Plane
**Portkey** is an AI gateway and control plane that lets teams route across multiple LLM providers, govern access, and observe usage through one API. By plugging Portkey in front of models like OpenAI, Anthropic, Google, Azure OpenAI, Amazon Bedrock, and local models, teams gain reliability features—**retries, fallbacks, caching, guardrails, and live analytics**—without refactoring app code. The company offers both an open-source gateway and a managed cloud control plane for org-wide governance.
Core value: **One fast gateway to multi‑LLM** with fine-grained key management and enterprise governance.Routing logic: **Conditional routing** on cost, latency, region, and quality signals, with **automatic failover** when providers throttle or go down.Governance and safety: **Redaction, policy checks, budgets, and guardrails**.Observability: Request tracing, token counts, and cost tracking per app or tenant.Explore the product on the [website](https://portkey.ai), see the [AI Gateway features](https://portkey.ai/features/ai-gateway), and review [docs for conditional routing](https://portkey.ai/docs/product/ai-gateway/conditional-routing) and [multi‑tenant key management](https://portkey.ai/docs/guides/use-cases/multi-tenant-ai-feature). The open-source gateway is available on [GitHub](https://github.com/Portkey-AI/gateway).
---
Why Portkey for AI Agents and LLM Apps
**Multi‑LLM routing + key management:** Bring model credentials once, issue virtual keys to teams/tenants, set routing and rate limits, and attribute costs across providers.**Reliability at scale:** Retries, fallbacks, and caching to improve uptime and latency; see guidance on [failover strategies](https://portkey.ai/blog/failover-routing-strategies-for-llms-in-production).**Production governance:** Budgets, quotas, policy enforcement, and guardrails for safety/compliance.**Unified observability:** End‑to‑end logs, traces, token, and spend analytics.Best for teams who need **reliability, governance, and cost control** across many models and applications.
---
Key Capabilities
**Routing and Reliability**Conditional routing by cost, latency, region, provider policy, or content rules**Automatic failover** and load balancing across providersCaching, rate limits, and usage caps**Governance and Guardrails**Virtual keys per team/tenant with quotas and budget controlsInput/output checks, redaction, and policy enforcementOrganization‑wide access controls and auditability**Observability**Centralized logging, tracing, token usage, and cost attributionPer‑use‑case and per‑tenant analytics for finance and ops**Deployment Choices**Managed cloud control planeSelf‑host using the [open‑source gateway](https://github.com/Portkey-AI/gateway) and DockerListings on [AWS Marketplace](https://aws.amazon.com/marketplace/pp/prodview-o2leb4xcrkdqa) and [Microsoft Marketplace](https://marketplace.microsoft.com/en-us/product/saas/portkey.enterprise-saas)---
Who It’s For
Product and platform teams shipping LLM features to productionEnterprises needing **governance, cost attribution, and policy controls**Startups choosing a managed gateway instead of maintaining a proxy in‑houseMulti‑tenant SaaS platforms requiring **per‑tenant keys, quotas, routing, and spend tracking**---
Common Use Cases
**Multi‑LLM routing** by cost, latency, region, or content policy**Failover and load balancing** to reduce downtime**AI cost management** with budgets and per‑use‑case attribution**Multi‑tenant key management** with virtual keys, rate limits, and usage caps**Guardrails** for safety and compliance checks**Observability** for requests, tokens, and spend---
Integrations and Ecosystem
**Model providers:** OpenAI, Anthropic, Google, Azure OpenAI, Amazon Bedrock; community mentions for local models (e.g., Ollama)**SDKs and frameworks:**Vercel AI SDK community providerLangroid notesMCP server integration discussed by the community**Tooling:** Reported usage in CI/CD and GitHub Actions in customer testimonials---
Pricing
See the [pricing page](https://portkey.ai/pricing).Third‑party listings indicate a Free plan (with log limits) and a Business plan around $99/month; Enterprise via custom pricing and marketplaces. Reference sources: [SaaSworthy](https://www.saasworthy.com/product/portkey-ai/pricing), [Doctor Droid](https://drdroid.io/entries/portkey-ai-gateway), [AWS Marketplace](https://aws.amazon.com/marketplace/pp/prodview-o2leb4xcrkdqa).Conclusion: a **free tier exists**, and paid tiers scale by usage and feature set. Confirm specifics with sales.---
Customer Sentiment
Pros**Easy to integrate**—drop‑in for OpenAI‑compatible appsStrong **reliability features** (retries, fallbacks, caching) that improve cost and latencyHelpful support and onboardingGood managed alternative to self‑built gatewaysPositive enterprise reviews on production reliabilityConsSome teams prefer other gateways for specific “enterprisey” features or fully self‑hosted controlOccasional friction integrating local setups like Ollama (requires tuning/Docker)Mixed picks among Portkey, LiteLLM, TrueFoundry, Helicone depending on stack and deployment model---
Company Snapshot
Name: Portkey (Portkey Router)Industry: Software DevelopmentFounded: 2023Headquarters: San Francisco, USFounders: Rohit Agarwal (CEO), Ayush Garg (CTO)Backing: Lightspeed (per company profile)Team size: ~11–50 employees (26 listed on LinkedIn)Token volume claim: Processes 14B+ LLM tokens/day (per [LinkedIn company description](https://www.linkedin.com/company/portkey-ai))Official links:[Website](https://portkey.ai)[Docs](https://portkey.ai/docs)[LinkedIn](https://www.linkedin.com/company/portkey-ai)[X/Twitter](https://x.com/portkeyai)[GitHub](https://github.com/portkey-ai)[Open‑source gateway](https://github.com/Portkey-AI/gateway)Leadership profiles: [Rohit Agarwal](https://www.linkedin.com/in/1rohitagarwal), [Ayush Garg](https://www.linkedin.com/in/ayush92garg)
---
Deployment and Security Notes
Deployment options: Managed cloud, self‑host via Docker, and open‑source gateway. Available on [AWS Marketplace](https://aws.amazon.com/marketplace/pp/prodview-o2leb4xcrkdqa) and [Microsoft Marketplace](https://marketplace.microsoft.com/en-us/product/saas/portkey.enterprise-saas).Security/compliance: Governance and guardrails are core features. No public SOC 2 page identified—enterprises should **confirm certifications, data handling, and isolation details with sales**.---
Fast Facts and Highlights
Product focus: **Multi‑LLM routing, key management, retries, caching, guardrails, observability, governance**Positioning: “Production stack for GenAI builders” with gateway, observability, guardrails, and governanceOpen source: Routes to 200+ LLMs and supports 50+ guardrails via one API (per [gateway repo](https://github.com/Portkey-AI/gateway))---
Additional Resources
[AI Gateway features](https://portkey.ai/features/ai-gateway)[Conditional routing docs](https://portkey.ai/docs/product/ai-gateway/conditional-routing)[Failover strategies](https://portkey.ai/blog/failover-routing-strategies-for-llms-in-production)[Multi‑tenant guide](https://portkey.ai/docs/guides/use-cases/multi-tenant-ai-feature)[Pricing](https://portkey.ai/pricing)[AWS Marketplace listing](https://aws.amazon.com/marketplace/pp/prodview-o2leb4xcrkdqa)[Microsoft Marketplace listing](https://marketplace.microsoft.com/en-us/product/saas/portkey.enterprise-saas)[G2 reviews](https://www.g2.com/products/portkey/reviews)
