# LiteLLM review

> LiteLLM is an open source AI gateway and Python SDK that calls 100+ LLM providers through one OpenAI-format interface.

Devtoolwatch Board, the week of Oct 25–31, 2026. Canonical page: https://index.devtoolwatch.com/tools/litellm/

## TL;DR

LiteLLM gives Python developers and platform teams one OpenAI-format interface to 100+ LLM providers, including OpenAI, Anthropic, Gemini, Bedrock and Azure. You can use it as a Python SDK or self-host it as a proxy server. The proxy adds virtual keys, spend tracking, guardrails, load balancing, MCP and A2A gateways, and an admin dashboard.

## LiteLLM at a glance

- **Category:** [Model SDKs and gateways](https://index.devtoolwatch.com/categories/model-sdks-and-gateways/)
- **Maintained by:** Berri AI (organization, [@BerriAI](https://github.com/BerriAI) on GitHub)
- **Language:** Python, TypeScript, Rust
- **Repository:** [github.com/BerriAI/litellm](https://github.com/BerriAI/litellm)
- **Website:** [docs.litellm.ai/docs](https://docs.litellm.ai/docs/)
- **PyPI package:** [litellm](https://pypi.org/project/litellm/) 1.106.2
- **Latest release:** [v1.105.3](https://github.com/BerriAI/litellm/releases/tag/v1.105.3) on GitHub, Oct 21, 2026
- **First published:** Jul 27, 2023

## What LiteLLM does

- LiteLLM calls 100+ LLM providers with the same completion() interface and a consistent output format.
- LiteLLM supports endpoints such as /chat/completions, /responses, /embeddings, /images, /audio, /batches, /rerank and /messages.
- LiteLLM Proxy adds virtual keys, spend tracking, guardrails, load balancing and an admin dashboard for teams.
- LiteLLM includes retry and fallback logic across multiple deployments through its Router.
- LiteLLM connects MCP servers to any LLM, either through the SDK bridge or the gateway's /chat/completions endpoint.
- LiteLLM invokes A2A agents, including LangGraph, Vertex AI Agent Engine, Azure AI Foundry, Bedrock AgentCore and Pydantic AI agents.
- LiteLLM runs Claude Code, Codex, OpenCode, pi or Deep Agents on any model through its Python SDK agent harnesses.

### How it works

LiteLLM can be imported as a Python library, and your code calls completion() with provider-prefixed model names. Teams can also run it as a proxy server, either from the CLI or as a container. Clients then point an OpenAI client at the proxy's base URL. The proxy forwards requests to providers like OpenAI, Anthropic, Bedrock or Azure. It can also expose MCP tools and A2A agents.

## How to install LiteLLM

**Python SDK with uv**

```
uv add litellm
```

**Proxy server with uv**

```
uv tool install 'litellm[proxy]'
```

**Quickstart script (macOS and Linux)**

```
curl -fsSL https://raw.githubusercontent.com/BerriAI/litellm/main/scripts/quickstart.sh | sh
```

### Requirements

- Python 3.10 or later, below 3.15
- API keys for each LLM provider you call
- Git, uv and the Rust build toolchain, only if you build litellm-core from source

### First steps

1. Set provider keys such as OPENAI_API_KEY or ANTHROPIC_API_KEY as environment variables.
2. Call completion() with a model name like openai/gpt-4o and a messages list.
3. To use the proxy, run `litellm --model gpt-4o` to start the gateway.
4. Point an OpenAI client at base_url http://0.0.0.0:4000 to send requests through the proxy.

## Who LiteLLM is for

- Python developers who want to switch LLM providers without rewriting code.
- Platform teams running a self-hosted, centralized LLM gateway with keys and spend tracking.
- Teams that need governed access to MCP tools and A2A agents across models.

### Limitations

- LiteLLM's litellm and litellm-core packages cannot be installed in the same environment because they share Python files.
- LiteLLM-core has no CLI entry points, optional extras or bundled dashboard. Use litellm for the proxy.
- LiteLLM versions 1.82.7 and 1.82.8 on PyPI were reported compromised with a credential-stealing file.
- LiteLLM's MCP OAuth can fail when providers refuse dynamic client registration. A pre-registered OAuth app may be needed.

## Usage data (the week of Oct 25–31, 2026)

- Board rank: #101 overall, #3 in Model SDKs and gateways, momentum score 63
- Downloads that week: 23.1M (+4% vs 4-week average)
- New GitHub stars per week (last 12 weeks): Aug 9: 542; Aug 16: 618; Aug 23: 559; Aug 30: 596; Sep 6: 504; Sep 13: 610; Sep 20: 533; Sep 27: 454; Oct 4: 873; Oct 11: 581; Oct 18: 582; Oct 25: 519

## LiteLLM alternatives

| Rank | Tool | Stars | Downloads/wk |
|---|---|---|---|
| #1 | [thinking-orbs](https://index.devtoolwatch.com/tools/jakubantalik-thinking-orbs/) | 3.9k | 1.4M |
| #2 | [Headroom](https://index.devtoolwatch.com/tools/headroomlabs-ai-headroom/) | 78.5k | 125.8k |
| #4 | [Honcho](https://index.devtoolwatch.com/tools/plastic-labs-honcho/) | 7.9k | – |
| #5 | [Anthropic SDK](https://index.devtoolwatch.com/tools/anthropic-sdk/) | 4k | 111M |
| #6 | [Hugging Face Hub](https://index.devtoolwatch.com/tools/hf-hub/) | 4k | 68.3M |
| #7 | [OpenAI SDK](https://index.devtoolwatch.com/tools/openai-sdk/) | 32k | 141.3M |

## LiteLLM FAQ

### Is LiteLLM free and open source?

LiteLLM's code is on GitHub (BerriAI/litellm), but the repository doesn't declare a standard open-source license. Check its license file before using it commercially.

### How do I install LiteLLM?

Run `uv add litellm`. The litellm package needs Python 3.10 or newer, below 3.15. The install section above has the other options from the project's README.

### Is LiteLLM still maintained?

Yes. The BerriAI/litellm repository was last updated on Oct 21, 2026, and the latest release, v1.105.3 on GitHub, came out on Oct 21, 2026.

### How popular is LiteLLM?

LiteLLM has 62.5k GitHub stars as of Nov 1, 2026. It gained 519 new stars in the week of Oct 25–31, 2026, against a 12-week average of 581 a week. Its main package had 23.1M downloads that week, +4% against its own 4-week average. That made it #3 among LLM SDKs and gateways on the Devtoolwatch Board.

### LiteLLM vs Bifrost: how do they compare?

LiteLLM and Bifrost are both tracked by Devtoolwatch. LiteLLM has 62.5k GitHub stars and Bifrost has 9.3k. [Bifrost review](https://index.devtoolwatch.com/tools/maximhq-bifrost/)

### LiteLLM vs vLLM: how do they compare?

LiteLLM and vLLM are both tracked by Devtoolwatch. LiteLLM has 62.5k GitHub stars and vLLM has 96k. In the week of Oct 25–31, 2026, LiteLLM's main package had 23.1M downloads and vLLM's had 677.2k. vLLM is an open-source Python library and engine for high-throughput, memory-efficient inference and serving of large language models. [vLLM review](https://index.devtoolwatch.com/tools/vllm/)

### What are the best alternatives to LiteLLM?

Other LLM SDKs and gateways on the Devtoolwatch Board in the week of Oct 25–31, 2026, by momentum: thinking-orbs (#1), Headroom (#2), Honcho (#4), Anthropic SDK (#5), Hugging Face Hub (#6). The alternatives table above has their numbers.

### Is LiteLLM an AI gateway?

Yes. LiteLLM describes itself as an open source AI gateway for 100+ LLMs. Its proxy server provides virtual keys, spend tracking, guardrails, load balancing and an admin dashboard. It also includes MCP and A2A agent gateways.

### Is LiteLLM safe to use?

LiteLLM signs its Docker images with cosign, and you can verify them against a pinned key. Separately, PyPI versions 1.82.7 and 1.82.8 were reported compromised with a malicious litellm_init.pth credential stealer. Avoid those versions.

### How do I use the LiteLLM proxy?

Install the proxy with `uv tool install 'litellm[proxy]'` and start it with `litellm --model gpt-4o`. Then create an OpenAI client with base_url http://0.0.0.0:4000 and send chat completion requests as usual.

### How do I use LiteLLM with Claude Code?

LiteLLM's Python SDK runs Claude Code through litellm.agent with Harness.CLAUDE_CODE. Set LITELLM_PROXY_API_BASE and LITELLM_PROXY_API_KEY to route the agent's model calls through your gateway. You also need starlette, uvicorn and the claude CLI installed.

### Can I use LiteLLM with Cursor?

Yes. LiteLLM's README shows a Cursor mcpServers config that points to http://localhost:4000/mcp/. The config passes your master key in the x-litellm-api-key header.

## Sources

1. [BerriAI/litellm on GitHub](https://github.com/BerriAI/litellm)
2. [LiteLLM README](https://github.com/BerriAI/litellm/blob/main/README.md)
3. [litellm on PyPI](https://pypi.org/project/litellm/)
4. [LiteLLM website](https://docs.litellm.ai/docs/)
5. [Release v1.105.3 (Oct 21, 2026)](https://github.com/BerriAI/litellm/releases/tag/v1.105.3)
6. [Release v1.106.2 (Oct 21, 2026)](https://github.com/BerriAI/litellm/releases/tag/v1.106.2)
7. [Release v1.103.7 (Oct 21, 2026)](https://github.com/BerriAI/litellm/releases/tag/v1.103.7)
8. [Hacker News: Tell HN: Litellm 1.82.7 and 1.82.8 on PyPI are compromised](https://news.ycombinator.com/item?id=47501426)
9. [Hacker News: Malicious litellm_init.pth in litellm 1.82.8 PyPI package – credential stealer](https://news.ycombinator.com/item?id=47501729)
10. [Hacker News: Show HN: liteLLM Proxy Server: 50+ LLM Models, Error Handling, Caching](https://news.ycombinator.com/item?id=37095542)
11. [Devtoolwatch Board, the week of Oct 25–31, 2026](https://index.devtoolwatch.com/)