Overview

Routerly sits between applications and AI providers, giving developers a shared point for routing requests and tracking usage. It supports OpenAI and Anthropic wire formats, including streaming, so compatible clients can connect by changing the base URL. Supported providers include OpenAI, Anthropic, Google Gemini, Ollama, Mistral, Cohere, xAI, and custom HTTP endpoints.

Projects can have separate Bearer tokens, model pools, routing configurations, and budget limits. Routerly stores configuration and usage records in JSON files rather than requiring an external database. Its maker says prompts and API keys stay on the user's infrastructure; provider API keys and project tokens are AES-256 encrypted at rest, while user passwords are bcrypt-hashed.

Key features

Policy-based routing

Up to nine configurable policies cover model selection, cost, provider health, performance, capability, context, budget, rate limits, and fairness. This breadth gives teams room to shape routing around operational priorities, while also making Routerly more than a simple pass-through. Fallback routing is supported for directing requests elsewhere when needed.

Cost tracking and budgets

Per-token cost tracking uses built-in pricing presets or custom model pricing. Budgets can apply globally, per project, or per token, and measure cost, calls, or token usage across rolling or calendar windows. That granularity is useful for keeping separate projects within limits, though teams must configure their own gateway and model pricing.

Client compatibility

The OpenAI and Anthropic wire formats and streaming support let compatible applications connect through a base-URL change. Routerly names Cursor, Open WebUI, OpenClaw, LangChain, and LlamaIndex among the clients and frameworks that can connect to it.

Pricing

Free self-hosted: 0.00 USD per free. The plan is self-hosted, licensed under AGPL-3.0, and adds no markup to API calls. It is a practical fit for developers and teams willing to operate their own gateway; there is no paid hosted tier or trial described in this offer. Provider API usage remains subject to the providers themselves.

The trade-off is operational responsibility: Docker and installer-based deployment are provided, but Routerly is not a managed service. The installer supports Windows, macOS, and Linux.

Platforms

Routerly supports API, Linux, macOS, self-hosted, web, and Windows. Docker and installer-based deployment give teams multiple ways to run it, including on the major desktop operating systems.

Who it's for

Routerly suits developers, SaaS teams, and local-first projects that want configurable provider routing, cost tracking, and budget controls without handing prompt traffic to a hosted gateway. Its project isolation can help teams separate credentials and limits. It is a poor fit for buyers seeking a managed gateway or those unwilling to deploy and maintain infrastructure.

Pros and cons

Pros: Up to nine routing policies and fallback routing give teams meaningful control over request handling.
Pros: Project-specific tokens, model pools, routing, and budgets support separation between workloads.
Pros: JSON storage avoids an external database dependency, and the self-hosted AGPL-3.0 offer adds no markup to API calls.
Cons: Self-hosting means deployment and operation remain the user's responsibility.
Cons: Cost visibility depends on configuring pricing presets or custom model prices to match the models in use.

Alternatives

For more options, browse LLM Gateway Software.

YoloRouter is another free self-hosted option; choose it if a single binary with SQLite out of the box and optional PostgreSQL better suits your deployment.
Kong Gateway has an open-source gateway at 0.00 USD per free and a Konnect free trial; consider it if you want an API gateway rather than a gateway centered on AI provider routing.
Apache APISIX is a free, self-hosted API gateway option.
Manifest offers unlimited self-hosted use on its Free plan, alongside cloud routing capped at 10,000 requests per month; consider it if those cloud request terms fit better.
OpenRouter offers hosted access to 25+ free models from four free providers, with 50 requests per day and five workspaces on Free; pick it if those hosted model options are preferable to operating your own gateway.
Requesty offers 200 requests per day on Free, including routing, caching, fallbacks, spend tracking, analytics, and EU data residency; consider it if a hosted free tier and daily request allowance suit your needs.
Helicone is a freemium, open-source alternative for API, self-hosted, and web use.
LiteLLM offers a free, self-hosted open-source plan with 100+ providers, virtual keys, spend tracking, budgets, and rate limits; consider it if that stated provider breadth and feature set is the better match.

Verdict

Choose Routerly if you want an infrastructure-controlled gateway with flexible provider routing, per-token cost visibility, and budgets that can be scoped to projects and tokens. Its free, no-markup, AGPL-3.0 self-hosted offer is compelling for developers prepared to run it themselves. Look elsewhere if you need a managed service rather than another system to deploy and operate.

	Eden AI	Routerly
Free plan	No	Yes
Free trial	No	No
Paid from	—	Free
Open source	No	No
Platforms	api, Web	api, Linux, macOS, self-hosted, Web, Windows
Free plan	No	Yes
Fallback routing	Yes	Yes
Usage analytics	Yes	Yes
Self-hosted deployment	No	Yes
Budget controls	Yes	Yes
Virtual API keys	Yes	Yes

Eden AI vs Routerly

Eden AI

Routerly