Routerly

What Routerly is
Routerly is a self-hosted gateway that routes application requests to AI providers while tracking costs and enforcing budgets. Its routing policies can account for model choice, cost, provider health and performance, capabilities, context, budgets, rate limits, and fairness. It accepts OpenAI and Anthropic wire formats, including streaming, so compatible clients can connect by changing the base URL. Listed providers include OpenAI, Anthropic, Google Gemini, Ollama, Mistral, Cohere, xAI, and custom HTTP endpoints. Cost tracking works per token with built-in pricing presets or custom model prices. Budgets can apply globally, per project, or per token, and measure cost, calls, or token use over rolling or calendar windows. Projects can separate tokens, model pools, routing settings, and budget limits. Routerly stores configuration and usage records in JSON files without requiring an external database. It can be deployed with Docker or an installer. The software is free to self-host under AGPL-3.0, and its maker says it adds no markup to API calls.
Who it suits
Routerly is aimed at developers, including SaaS teams and local-first projects, who want to manage AI-provider routing and budgets on infrastructure they control.
What it does well
- Supports OpenAI and Anthropic wire formats with streaming.
- Offers configurable routing policies.
- Sets budgets by project or token.
- Stores records without an external database.
- Free self-hosting under AGPL-3.0.
What to know first
- Requires self-hosting.
- No free trial is listed.
Limits
The free self-hosted plan is licensed under AGPL-3.0. The listed plan has no markup on API calls.
Plans
| Plan | Billed and limits | Price |
|---|---|---|
| Free self-hosted | Free · Self-hosted · No markup on API calls · AGPL-3.0 | Free |
As routerly.ai lists them, read 30 Sept 2026. The maker's own page has the final word.
Getting started
- Visit the Routerly website.
- Choose Docker or installer-based deployment.
- Install on Windows, macOS, or Linux using the installer, or deploy with Docker.
- Connect a compatible client using the OpenAI or Anthropic wire format and Routerly base URL.
- Configure providers, routing policies, and budgets.
Routerly is a self-hosted gateway for developers and teams that need to manage traffic across AI providers. It suits projects that value infrastructure control and per-project budgets; its main trade-off is that you operate the gateway yourself.
Overview
Routerly sits between client applications and AI services, giving teams one place to route requests, monitor costs, and apply usage limits. OpenAI- and Anthropic-compatible wire formats, including streaming, let compatible clients connect by changing the base URL. It supports OpenAI, Anthropic, Google Gemini, Ollama, Mistral, Cohere, xAI, and custom HTTP endpoints.
Projects can use separate Bearer tokens, model pools, routing settings, and budget limits. Routerly stores configuration and usage records in JSON files without an external database. Its maker says prompts and API keys remain on the user's infrastructure; provider keys and project tokens are AES-256 encrypted at rest, and passwords are bcrypt-hashed.
Key features
Policy-based routing
Up to nine policies cover model selection, cost, provider health, performance, capability, context, budget, rate limits, and fairness. That range is useful when routing must reflect several operational priorities, while adding more control than a simple provider pass-through. Fallback routing is supported, too.
Cost and budget controls
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 over rolling or calendar windows. This makes it easier to set limits at different levels, though the team still needs to run and configure its own gateway.
Client compatibility and deployment
Routerly names Cursor, Open WebUI, OpenClaw, LangChain, and LlamaIndex among the clients and frameworks that can connect. Docker and installer-based deployment are available; the installer supports Windows, macOS, and Linux. The API, Linux, macOS, self-hosted, web, and Windows platform range gives developers multiple ways to fit it into their environment.
Pricing
Free self-hosted: free. The plan is self-hosted, carries no markup on API calls, and is licensed under AGPL-3.0. There is no free trial because the self-hosted plan is free. Provider usage remains subject to the providers' own charges; Routerly's no-markup policy does not make inference free.
There are no paid tiers or published seat, request, or usage quotas to weigh against the free plan. Its cost advantage comes with the responsibility of deploying and operating the gateway yourself. That trade-off suits developers and teams seeking control; it is less compelling for readers who want a managed service rather than self-hosting.
Who it's for
Routerly is a strong fit for developers, SaaS teams, and local-first projects that want a shared gateway with provider choice, project isolation, and configurable spending controls. The combination of self-hosting, broad routing policies, and no external database requirement also suits teams that prefer to keep prompts and keys on their own infrastructure. Readers who would rather avoid operating software should look at hosted alternatives.
Pros and cons
- Pros: Broad routing policies and fallback support let teams shape provider selection around cost, health, capability, and other criteria.
- Pros: Project-level tokens, model pools, and budgets help separate workloads without requiring a database service.
- Pros: The free AGPL-3.0 plan adds no markup to API calls, and the maker says prompts and keys stay on the user's infrastructure.
- Cons: Self-hosting means deployment and operation are the user's responsibility, rather than a managed gateway handling that work.
- Cons: AGPL-3.0 may not suit teams that require a different licensing arrangement.
Alternatives
LLM Gateway Software is a useful starting point for comparing options across the category. Choose Bifrost if its free-forever OSS plan and Apache 2.0 license better fit your deployment or licensing needs; it also offers npx, Docker, Kubernetes, and Go binary deployment. YoloRouter is another free self-hosted option, with a single binary and SQLite out of the box plus optional PostgreSQL.
Kong Gateway may fit readers looking for an open-source API gateway with a separate Konnect free trial. Helicone is an alternative to consider if its 10,000-request Hobby plan is a closer match. LiteLLM may suit teams looking for its stated 100+ provider coverage and self-hosted virtual keys, users, teams, spend tracking, budgets, and rate limits. Manifest offers a free plan with unlimited self-hosted use, or 10,000 routed requests per month on Cloud. For a hosted option, OpenRouter has a free plan with 25+ free models, four free providers, five workspaces, and 50 requests a day. Requesty is another hosted choice, with a free plan that includes 200 requests per day, routing, caching, fallbacks, spend tracking, analytics, and EU data residency.
Verdict
Choose Routerly if you want a free, self-hosted gateway that combines detailed routing controls with per-project isolation and flexible budgets. The decisive reason to look elsewhere is the same one that defines its appeal: Routerly expects you to deploy and run it yourself.
Verdict
Routerly brings provider routing, cost tracking, and budget controls into a self-hosted gateway. It is free under AGPL-3.0 and adds no markup to API calls.
Questions
How much does Routerly cost?
The Free self-hosted plan costs nothing. The site says Routerly adds no markup to API calls.
Is there a free trial?
No. Routerly is listed as free, with no free trial.
Which providers does Routerly support?
Supported providers include OpenAI, Anthropic, Google Gemini, Ollama, Mistral, Cohere, xAI, and custom HTTP endpoints.
Can clients connect using OpenAI or Anthropic formats?
Yes. Routerly supports both wire formats, including streaming; compatible clients can connect by changing the base URL.
Is Routerly open source?
The free self-hosted software is licensed under AGPL-3.0.
Where does Routerly run?
Listed platforms include API, Linux, macOS, self-hosted, web, and Windows. The installer supports Windows, macOS, and Linux.
Facts from Routerly's own pages
- API compatibility
- It supports OpenAI and Anthropic wire formats with streaming, so clients can use Routerly by changing the base URL. · routerly.ai, 30 Sept 2026
- Budget controls
- Budgets can be set globally, per project, or per token, and can track cost, calls, or token usage over rolling or calendar windows. · doc.routerly.ai, 30 Sept 2026
- Cost tracking
- Routerly tracks per-token costs and supports built-in pricing presets and custom model pricing. · routerly.ai, 30 Sept 2026
- Data handling
- Routerly is self-hosted, and its maker says prompts and API keys stay on the user's infrastructure. · blog.routerly.ai, 30 Sept 2026
- Deployment
- The maker provides Docker and installer-based deployment, and says the installer supports Windows, macOS, and Linux. · routerly.ai, 30 Sept 2026
- Integrations
- The site names Cursor, Open WebUI, OpenClaw, LangChain, and LlamaIndex as clients or frameworks that can connect to Routerly. · routerly.ai, 30 Sept 2026
- Intended users
- The site describes Routerly as built for developers and use cases ranging from indie projects to enterprise deployments, including SaaS teams and local-first development. · routerly.ai, 30 Sept 2026
- License and pricing
- Routerly is free to self-host under the AGPL-3.0 license, and the site says it adds no markup to API calls. · routerly.ai, 30 Sept 2026
- Product
- Routerly is a self-hosted gateway between applications and AI providers that routes requests, tracks costs, and enforces budgets. · routerly.ai, 30 Sept 2026
- Project isolation
- Each project can have separate Bearer tokens, model pools, routing configuration, and budget limits. · routerly.ai, 30 Sept 2026
- Providers
- Supported providers include OpenAI, Anthropic, Google Gemini, Ollama, Mistral, Cohere, xAI, and custom HTTP endpoints. · routerly.ai, 30 Sept 2026
- Routing
- Its routing engine uses up to nine configurable policies, including LLM selection, cost, health, performance, capability, context, budget, rate limits, and fairness. · blog.routerly.ai, 30 Sept 2026
- Security
- The documentation says provider API keys and project tokens are AES-256 encrypted at rest, and user passwords are bcrypt-hashed. · doc.routerly.ai, 30 Sept 2026
- Storage and dependencies
- Routerly stores configuration and usage records in JSON files and does not require an external database. · doc.routerly.ai, 30 Sept 2026
Where Routerly stands on Freedom251
- LLM gateways#10 of 33
- LLM Gateway Software#14 of 34