18.7
Manifest is an open-source LLM gateway for routing model requests, recovering from provider problems, and tracking inference spending.
Free planOpen sourceWebAPILinux
Best LLM Gateways in 2026
LLM gateways provide a software layer for working with language-model APIs. They can run through an API or web service, or on your own infrastructure, depending on the product. Phone users should verify Android support on each product page; these notes otherwise cover API, web, self-hosted, Linux, macOS and Windows.
28.6
LangSmith is a platform for building, evaluating, deploying, monitoring, and improving AI agents.
Free planWebAPILinux
38.3
Bifrost is an AI gateway and control plane for routing, governing, and securing traffic involving model calls, MCP tools, and agents.
Free planOpen sourceWebAPILinux
- 4
LLM Gateway
LLM Gateway gives product and platform teams one OpenAI-compatible API for routing requests across more than 250 models from over 40 providers.
Free planOpen sourceWebAPIExtensionLinux8.2Free plan - 58.1$5.50 one-time
- 6
TrueFoundry AI Gateway
TrueFoundry AI Gateway sits between applications and model providers, giving teams a single API for managing model access, routing, governance,…
Free planWebAPILinuxSelf-hosted8.0From $499 a month - 7
YoloRouter
YoloRouter is an OpenAI-compatible gateway that routes AI model requests across multiple providers.
Free planOpen sourceWebAPILinuxmacOS7.8Free plan - 8
Kong Gateway
Kong Gateway is an open-source API gateway designed for microservices and distributed architectures in hybrid and multi-cloud environments.
Free planWebAPILinuxSelf-hosted7.5Free plan - 9
Helicone
Helicone is a platform for routing, debugging, and analyzing AI applications.
Free planWebAPISelf-hosted7.5From $79 a month - 10
TensorZero
TensorZero is an open-source stack for building and operating applications that use large language models.
Free planWebAPILinuxSelf-hosted7.4Free plan - 11
Requesty
Requesty is a managed AI gateway for connecting applications to more than 600 models through one OpenAI-compatible API.
Free planWebAPIExtension7.4Free plan - 12
Vercel
Vercel is infrastructure for building, deploying, securing, observing, and scaling apps and agents.
Free planWebAPI7.3$20 usage-based - 13
Envoy AI Gateway
Envoy AI Gateway, now named Agent Router, is an open-source control plane for managing AI and agent traffic using Envoy.
Free planAPILinuxmacOSSelf-hosted7.2Free plan - 14
UniterRouter
UniterRouter is a unified AI gateway available through an API and web interface.
Free planWebAPI7.1From $500 a year - 15
OpenRouter
OpenRouter provides access to hundreds of AI models through a single API endpoint.
Free planWebAPI7.0Free plan - 16
Routerly
Routerly is a self-hosted gateway that routes application requests to AI providers, tracks usage costs, and enforces budgets.
Free planWebAPILinuxmacOS7.0Free plan - 17
ZenMux
ZenMux provides one API and API key for accessing AI models from OpenAI, Anthropic, Google, DeepSeek, and other providers.
Free planWebAPI7.0From $20 a month - 18
LiteLLM
Free planWebAPISelf-hosted6.8Free plan - 19
SGLang
SGLang is a free, open-source inference framework for serving language, vision-language, and diffusion models.
Free planAPILinuxmacOSSelf-hosted6.8Free plan - 20
Cloudflare AI Gateway
Free planWebAPI6.8Free plan - 21
Orq.ai
Orq.ai is an AI gateway and platform for routing, governing, and observing requests to models and tools.
Free planWebAPISelf-hosted6.7Free plan - 22
WSO2 AI Gateway
WebAPILinuxmacOS6.7From $119 a month - 23
Polyrouter
Polyrouter is a free, open-source LLM gateway that provides one endpoint for multiple model providers.
Free planWebAPISelf-hosted6.6Free plan - 24
Tyk AI Studio Proxy
Tyk AI Studio Proxy is an open-source control plane for managing organizational AI traffic, including LLMs, agents, MCP tool chains, and RAG workloads.
Free planWebAPISelf-hosted6.6Free plan - 25
Ramp Router
Ramp Router provides one API endpoint for sending requests to AI models from OpenAI, Anthropic, other providers, and selected open-source sources.
Free planWebAPI6.4Free plan
Choosing well
Compare plan terms
Freemium pricing is shown for LLM Gateway, Manifest, Bifrost, Helicone, LangSmith, LiteLLM and OpenRouter; TensorZero and Envoy AI Gateway are marked free.
Published paid entry points differ: Manifest starts at $19 per month, LangSmith at $39 per user per month and Helicone at $79 per month; Kong Gateway asks buyers to request pricing.
Check trial periods
Use a trial to evaluate a gateway before committing to payment. LLM Gateway and LiteLLM offer 30-day periods, Bifrost offers 14 days and Helicone offers seven.
Confirm how a trial changes to a paid plan and whether a free plan remains available after it ends.
Match your platform
Across the list, deployment options include API, web, self-hosted, Linux, macOS and Windows. Choose a gateway whose deployment options fit your server, desktop or browser setup.
Orq.ai is shown with API, self-hosted and web options. Check the product page for device support before relying on a phone workflow.
Review the code licence
Some tools in this category provide open-source code. When code access or hosting matters, review the repository, licence and deployment instructions before choosing.
Questions people ask
Are free plans available from LLM gateways?
Free plans are shown for tools including LLM Gateway, Manifest, TensorZero, Bifrost, Helicone, LangSmith, LiteLLM and OpenRouter.
Which gateways list trial periods?
LLM Gateway, LiteLLM, Bifrost and Helicone list trials lasting 30, 30, 14 and seven days, respectively.
Can an LLM gateway run on Android?
The platform information covers API, web, self-hosted, Linux, macOS and Windows. Check each product page for Android support before choosing.
Is self-hosting supported?
Orq.ai includes self-hosted deployment in its listing. Review each other product's platform and deployment details because their setups may differ.
Do any gateways provide open-source code?
Some tools in this category provide open-source code. Review the licence and hosting instructions if code access matters to your setup.