T

LLM gateways

TensorZero: Features, Integrations, and Pricing

tensorzero.com · checked 7 Oct 2026

TensorZero is a free, open-source stack for routing and managing large language model requests.

How it plugs in

—Public API
NoOpen source
YesFree tier

Signal on LLM gateways7.2#6 of 33

TensorZero brings an LLM gateway together with tools for tracking requests, testing changes, and using feedback to guide optimization. It connects to several model providers and can run on your own infrastructure. The project remains available on GitHub, but development has stopped and its repository is archived; teams should consider that status before adopting it.

TensorZero home page

Review

In a line

TensorZero combines an open-source LLM gateway with tools for working with request data and experiments. The project is no longer maintained.

What it does

A single gateway API connects to multiple model providers and supports routing, retries, and fallbacks. It also provides prompt templates, request schemas, and multi-step workflows.

Inference and feedback records go into a database you configure. Teams can use that data for monitoring, experiments, and optimization, including A/B tests that compare different options.

How it connects

The documented integrations cover Anthropic, AWS Bedrock, Azure, Google AI Studio, OpenAI, OpenRouter, Together AI, vLLM, and xAI. Other supported connections include OpenAI-compatible APIs, the OpenAI SDK, OpenTelemetry, and clients for major programming languages.

The project uses the Apache 2.0 License and is open source. Its gateway and interface can run through Docker, Docker Compose, Kubernetes and Helm, or a source build. TensorZero keeps data in a database you control, accessible with SQL.

Good fit

Teams seeking cross-provider routing, A/B testing, and control of their inference data may find TensorZero relevant. It includes fallback routing, budget controls, and virtual API keys.

Its self-hosted setup may also appeal to teams that want to manage deployment themselves.

Watch for

TensorZero is no longer maintained, and its GitHub repository is archived and read-only. Its maintenance status is therefore an important consideration before adoption.

API key authentication can protect gateway endpoints other than status and health checks. TensorZero also collects pseudonymous aggregate usage analytics; its documentation says these exclude application data and inference inputs and outputs, and that collection can be disabled.

Pricing

TensorZero is free.

Verdict

TensorZero may work for teams that want to host a model-request gateway, compare workflows, and retain data in a database they control, provided they can handle maintenance themselves. Its provider connections and experimentation tools address several common model-management tasks.

Before adopting it as a new long-term dependency, assess the risks of an archived project that is no longer maintained.

Our take

For teams seeking a free gateway they can host, TensorZero offers provider routing, fallbacks, and experimentation tools. Keeping data in a database you control also gives teams direct access to request and feedback records. Its archived, read-only repository and lack of ongoing maintenance are the central concerns. Adopt it only if your team can accept that maintenance risk.

  • Free and open source under the Apache 2.0 License
  • Supports several model providers and OpenAI-compatible APIs
  • Can store data in your own database for direct SQL queries
  • Includes routing, retries, fallbacks, and A/B testing
  • Development has stopped
  • GitHub lists the repository as archived and read-only
  • Deployment requires running the gateway and interface on your own infrastructure
  • Pick it ifTeams that want to host model routing and experimentation themselves; Teams that want request data in a database they control
  • Skip it ifYou need an actively maintained project; Your team cannot take responsibility for self-hosted deployment

Pricing

The TensorZero LLMOps platform plan is free, open source, and self-hosted.

  • TensorZero LLMOps platform100% self-hosted · open-sourceFree

As github.com lists them, checked 7 Oct 2026.

Questions

Is TensorZero free?

Yes. The TensorZero LLMOps platform is offered free of charge and is open source and self-hosted.

Is TensorZero still maintained?

No. The project’s homepage says it remains available on GitHub but is no longer maintained; GitHub marks the repository archived and read-only.

Which model providers does TensorZero support?

Its documented integrations include Anthropic, AWS Bedrock, Azure, Google AI Studio, OpenAI, OpenRouter, Together AI, vLLM, and xAI. It also works with OpenAI-compatible APIs.

Can TensorZero run on my own infrastructure?

Yes. You can deploy the gateway and interface with Docker, Docker Compose, or Kubernetes and Helm, or build them from source.

Where does TensorZero keep its data?

It writes data to a database you configure, which you can query using SQL.

Does TensorZero collect usage analytics?

It gathers pseudonymous aggregate usage analytics. The documentation says these exclude application data and inference inputs and outputs, and collection can be disabled.