LiteLLM

A Python SDK and a proxy server in one repository, the proxy giving applications one OpenAI-shaped address, one set of keys and one budget in front of many model providers.

Our description, from the project's README, as of 2026-09-03.

Rendered 2026-10-28

Key facts

Fact	Answer
Type	Proxy / gateway
Layer	Model access and routing (main layer)
Forms	library, gateway
Runs on	Your machine and a container environment you run
Models run	Your machine, the model company's cloud (OpenAI, Anthropic, Gemini) and a cloud provider's model service (Bedrock, Azure)
Model on your machine	Yes How to set it up
Several providers in one session	Yes How to set it up
Several endpoints built in	Yes How to set it up
Local and cloud together	Yes How to set it up

See this product in the grid with the others: Hybrid on your own machine and Hybrid in a container environment you run.

Configure for a local model

Use a local model: How to set it up (docs.litellm.ai page, read 2026-10-22) Settings the guide names: LM_STUDIO_API_BASE, LM_STUDIO_API_KEY, lm_studio/.
Use a local and a cloud model together: How to set it up (docs.litellm.ai page, read 2026-10-22) Settings the guide names: model_list, litellm_params, config.yaml, --config.
Hold several endpoints in the product: How to set it up (docs.litellm.ai page, read 2026-10-22) Settings the guide names: litellm_params, api_base, api_key, api_version.

Works with

Claude Desktop: a self-hosted gateway such as LiteLLM can be the inference provider, and it must implement Anthropic's Messages API (claude.com)
AG2: LLM configuration lists LiteLLM (docs.ag2.ai)
LM Studio: provider with an lm_studio/ model prefix (docs.litellm.ai)
Claude Code: runs Claude Code on any model through its Python SDK (github.com)
Codex: runs Codex on any model through its Python SDK (github.com)
OpenCode: runs OpenCode on any model through its Python SDK (github.com)

Data and compliance

Compliance: Set by whoever runs it, for their own deployment.
Data: Set by whoever runs it, with telemetry not checked yet.
Model and where it runs: Run by its operator.
Details: What the project's own pages state, not a compliance determination.

What the maker's own pages said on the dates shown, not a compliance determination.

Sources

LiteLLM page: docs (read 2026-10-26)
LiteLLM page: lm studio (read 2026-10-20)
LiteLLM page: ollama (read 2026-10-19)
LiteLLM page: docker quick start (read 2026-10-22)
GitHub page: litellm (read 2026-10-19, 2026-10-20, 2026-10-22, 2026-10-24)
GitHub page: litellm (the product's own site, from our list of products; not read for these lines)

The readings behind this page

Our summary of the vendor's pages as of the dates shown on each line (2026-10-19 to 2026-10-26); confirm on the vendor's page.

Local models

LiteLLM's LM Studio page says it supports all LM Studio models, selected with a model string that starts with lm_studio/. (docs.litellm.ai page, read 2026-10-20)

Cloud models

LiteLLM provides a single interface to call over 100 LLM providers, including OpenAI, Anthropic, Gemini, Bedrock, and Azure, using the OpenAI format. (github.com page, read 2026-10-20)

Works with other products

LiteLLM supports running Claude Code, Codex, Deep Agents, or OpenCode on any model via the Python SDK. (github.com page, read 2026-10-19)

About the product

LiteLLM's README describes LiteLLM as an open source AI gateway for over 100 LLMs and calls it self-hosted. (github.com page, read 2026-10-24)
LiteLLM's Ollama page says it supports all Ollama models, lists a running Ollama server as a prerequisite, and every call in its model table sets its address to localhost. (docs.litellm.ai page, read 2026-10-19)
LiteLLM's documentation describes its proxy as a gateway you host yourself that speaks the OpenAI API, and shows two ways to start it: the LiteLLM command-line tool or Docker. (docs.litellm.ai page, read 2026-10-26)
LiteLLM's documentation describes LiteLLM as an open-source library with one interface for calling more than 100 LLMs, among them OpenAI, Anthropic, Vertex AI and Bedrock. (docs.litellm.ai page, read 2026-10-26)

Terms·Catalog home

Our reading of vendors' pages, as of the dates shown. Confirm on the vendor's page before relying on any line that cites it. To report an error write to report@crinaro.ai