Portkey AI
Summary
An AI gateway and LLMOps platform for routing, observing, securing, and governing production AI applications across models and providers.
<p>Portkey AI is a production AI infrastructure platform combining a universal AI gateway with observability, prompt management, guardrails, model access controls, and MCP governance. Developers can route requests across providers, configure fallbacks, retries, caching, load balancing, rate limits, and budgets, then inspect logs and traces from one control plane. Portkey also publishes an open-source gateway for local or self-hosted deployment, while its managed plans add team, security, retention, and enterprise deployment capabilities.</p>
Categories: AI Safety and Detection, Models, APIs and Infrastructure, Data Analysis and BI, Automation and Integrations.
Pricing
Free, Paid - Starts at $0 - $49/month; custom plans available
Features
- Universal AI gateway with routing across multiple model providers
- Automatic fallbacks, retries, load balancing, timeouts, and conditional routing
- Simple and semantic caching for repeat requests
- OpenTelemetry-compatible logs, traces, analytics, feedback, and cost visibility
- Deterministic, model-based, and partner guardrails for input and output controls
- Prompt templates, versioning, variables, playground, and API deployment
- Model and key management with budgets, rate limits, and access controls
- MCP Gateway for authentication, tool provisioning, access control, and observability
- Open-source gateway with local and self-hosted deployment options
- Managed SaaS, hybrid, VPC, and air-gapped enterprise deployment options
Use Cases
- Route production AI traffic across multiple LLM providers
- Improve reliability with fallbacks, retries, caching, and load balancing
- Monitor AI requests, latency, token usage, costs, and failures
- Apply safety policies and guardrails before and after model calls
- Manage provider keys, budgets, rate limits, and model access across teams
- Version and deploy reusable prompts
- Govern MCP servers and tool access for AI agents
- Self-host an AI gateway or deploy it within private infrastructure
Pros
- Combines gateway routing, observability, guardrails, prompts, and governance in one platform
- Supports multi-provider and multimodal AI workloads through a consistent API
- Open-source gateway is available for local or self-hosted use
- Includes reliability controls such as fallbacks, retries, caching, and load balancing
- Offers centralized MCP authentication, permissions, and logging
Cons
- Free plan is explicitly aimed at prototyping rather than production workloads
- Production plan has recorded-log limits and metered overages
- Enterprise security and private deployment capabilities require custom pricing
- Adds another infrastructure layer that teams must configure and operate
- LLM provider charges are separate from Portkey plan fees
Last verified: 2026-07-20.
Alternatives
- ChatGPT - OpenAI's versatile chatbot for conversation, writing, and tasks.
- OpenAI Codex - OpenAI Codex is a ChatGPT-powered coding agent for building, reviewing, refactoring, testing, and shipping software across app, IDE, CLI, and web surfaces.
- Dia Browser - Dia is an AI browser from The Browser Company focused on work context, tabs, connected tools, reports, meetings, profiles, and privacy controls.
- Cline - Open-source AI coding agent for VS Code, JetBrains, terminal, and SDK workflows with model-provider flexibility including OpenRouter.
- Zapier - Zapier is a no-code automation platform for connecting apps, building workflows, and adding AI agents to business processes.
- Fish Audio - An AI voice platform for expressive text-to-speech, voice cloning, speech-to-text, voice design, and real-time voice applications.
Frequently Asked Questions
What is Portkey AI?
Portkey AI is an AI gateway and LLMOps platform for routing, observing, securing, and governing AI requests across models and providers.
Does Portkey AI have a free plan?
Yes. The Developer plan is free forever and includes 10,000 recorded logs per month. Portkey positions it for prototyping, testing, and enterprise proof-of-concept work rather than production workloads.
How much does Portkey AI cost?
The Production plan is listed at US$49 per month with 100,000 recorded logs and US$9 overages per additional 100,000 requests up to 3 million. Enterprise pricing is custom, and model-provider inference costs are separate.
Can Portkey be self-hosted?
Yes. Portkey publishes an open-source AI gateway that can run locally or be self-hosted. Enterprise customers can also discuss hybrid, VPC, and fully air-gapped deployment options.
What does the Portkey MCP Gateway do?
It centralizes authentication, access control, credential injection, tool provisioning, guardrails, and request logging for MCP servers used by agents and MCP clients.
Internal Links
Official website: https://portkey.ai/
Canonical FindMyAI page: https://findmyai.xyz/tools/portkey-ai