{"info":{"author":"BerriAI","author_email":null,"bugtrack_url":null,"classifiers":["License :: OSI Approved :: MIT License","Programming Language :: Python :: 3","Programming Language :: Python :: 3.10","Programming Language :: Python :: 3.11","Programming Language :: Python :: 3.12","Programming Language :: Python :: 3.13","Programming Language :: Python :: 3.14","Programming Language :: Python :: 3.9"],"description":"<h1 align=\"center\">\n        \ud83d\ude85 LiteLLM\n    <\/h1>\n    <p align=\"center\">\n        <p align=\"center\">Call 100+ LLMs in OpenAI format. [Bedrock, Azure, OpenAI, VertexAI, Anthropic, Groq, etc.]\n        <\/p>\n        <p align=\"center\">\n        <a href=\"https:\/\/render.com\/deploy?repo=https:\/\/github.com\/BerriAI\/litellm\" target=\"_blank\" rel=\"nofollow\"><img src=\"https:\/\/render.com\/images\/deploy-to-render-button.svg\" alt=\"Deploy to Render\"><\/a>\n        <a href=\"https:\/\/railway.app\/template\/HLP0Ub?referralCode=jch2ME\">\n          <img src=\"https:\/\/railway.app\/button.svg\" alt=\"Deploy on Railway\">\n        <\/a>\n        <\/p>\n    <\/p>\n<h4 align=\"center\"><a href=\"https:\/\/docs.litellm.ai\/docs\/simple_proxy\" target=\"_blank\">LiteLLM Proxy Server (AI Gateway)<\/a> | <a href=\"https:\/\/docs.litellm.ai\/docs\/enterprise#hosted-litellm-proxy\" target=\"_blank\"> Hosted Proxy<\/a> | <a href=\"https:\/\/docs.litellm.ai\/docs\/enterprise\"target=\"_blank\">Enterprise Tier<\/a><\/h4>\n<h4 align=\"center\">\n    <a href=\"https:\/\/pypi.org\/project\/litellm\/\" target=\"_blank\">\n        <img src=\"https:\/\/img.shields.io\/pypi\/v\/litellm.svg\" alt=\"PyPI Version\">\n    <\/a>\n    <a href=\"https:\/\/www.ycombinator.com\/companies\/berriai\">\n        <img src=\"https:\/\/img.shields.io\/badge\/Y%20Combinator-W23-orange?style=flat-square\" alt=\"Y Combinator W23\">\n    <\/a>\n    <a href=\"https:\/\/wa.link\/huol9n\">\n        <img src=\"https:\/\/img.shields.io\/static\/v1?label=Chat%20on&message=WhatsApp&color=success&logo=WhatsApp&style=flat-square\" alt=\"Whatsapp\">\n    <\/a>\n    <a href=\"https:\/\/discord.gg\/wuPM9dRgDw\">\n        <img src=\"https:\/\/img.shields.io\/static\/v1?label=Chat%20on&message=Discord&color=blue&logo=Discord&style=flat-square\" alt=\"Discord\">\n    <\/a>\n    <a href=\"https:\/\/www.litellm.ai\/support\">\n        <img src=\"https:\/\/img.shields.io\/static\/v1?label=Chat%20on&message=Slack&color=black&logo=Slack&style=flat-square\" alt=\"Slack\">\n    <\/a>\n    <a href=\"https:\/\/codspeed.io\/BerriAI\/litellm?utm_source=badge\">\n        <img src=\"https:\/\/img.shields.io\/endpoint?url=https:\/\/codspeed.io\/badge.json\" alt=\"CodSpeed\"\/>\n    <\/a>\n<\/h4>\n\n<img width=\"2688\" height=\"1600\" alt=\"Group 7154 (1)\" src=\"https:\/\/github.com\/user-attachments\/assets\/c5ee0412-6fb5-4fb6-ab5b-bafae4209ca6\" \/>\n\n\n## Use LiteLLM for\n\n<details open>\n<summary><b>LLMs<\/b> - Call 100+ LLMs (Python SDK + AI Gateway)<\/summary>\n\n[**All Supported Endpoints**](https:\/\/docs.litellm.ai\/docs\/supported_endpoints) - `\/chat\/completions`, `\/responses`, `\/embeddings`, `\/images`, `\/audio`, `\/batches`, `\/rerank`, `\/a2a`, `\/messages` and more.\n\n### Python SDK\n\n```shell\npip install litellm\n```\n\n```python\nfrom litellm import completion\nimport os\n\nos.environ[\"OPENAI_API_KEY\"] = \"your-openai-key\"\nos.environ[\"ANTHROPIC_API_KEY\"] = \"your-anthropic-key\"\n\n# OpenAI\nresponse = completion(model=\"openai\/gpt-4o\", messages=[{\"role\": \"user\", \"content\": \"Hello!\"}])\n\n# Anthropic  \nresponse = completion(model=\"anthropic\/claude-sonnet-4-20250514\", messages=[{\"role\": \"user\", \"content\": \"Hello!\"}])\n```\n\n### AI Gateway (Proxy Server)\n\n[**Getting Started - E2E Tutorial**](https:\/\/docs.litellm.ai\/docs\/proxy\/docker_quick_start) - Setup virtual keys, make your first request\n\n```shell\npip install 'litellm[proxy]'\nlitellm --model gpt-4o\n```\n\n```python\nimport openai\n\nclient = openai.OpenAI(api_key=\"anything\", base_url=\"http:\/\/0.0.0.0:4000\")\nresponse = client.chat.completions.create(\n    model=\"gpt-4o\",\n    messages=[{\"role\": \"user\", \"content\": \"Hello!\"}]\n)\n```\n\n[**Docs: LLM Providers**](https:\/\/docs.litellm.ai\/docs\/providers)\n\n<\/details>\n\n<details>\n<summary><b>Agents<\/b> - Invoke A2A Agents (Python SDK + AI Gateway)<\/summary>\n\n[**Supported Providers**](https:\/\/docs.litellm.ai\/docs\/a2a#add-a2a-agents) - LangGraph, Vertex AI Agent Engine, Azure AI Foundry, Bedrock AgentCore, Pydantic AI\n\n### Python SDK - A2A Protocol\n\n```python\nfrom litellm.a2a_protocol import A2AClient\nfrom a2a.types import SendMessageRequest, MessageSendParams\nfrom uuid import uuid4\n\nclient = A2AClient(base_url=\"http:\/\/localhost:10001\")\n\nrequest = SendMessageRequest(\n    id=str(uuid4()),\n    params=MessageSendParams(\n        message={\n            \"role\": \"user\",\n            \"parts\": [{\"kind\": \"text\", \"text\": \"Hello!\"}],\n            \"messageId\": uuid4().hex,\n        }\n    )\n)\nresponse = await client.send_message(request)\n```\n\n### AI Gateway (Proxy Server)\n\n**Step 1.** [Add your Agent to the AI Gateway](https:\/\/docs.litellm.ai\/docs\/a2a#adding-your-agent)\n\n**Step 2.** Call Agent via A2A SDK\n\n```python\nfrom a2a.client import A2ACardResolver, A2AClient\nfrom a2a.types import MessageSendParams, SendMessageRequest\nfrom uuid import uuid4\nimport httpx\n\nbase_url = \"http:\/\/localhost:4000\/a2a\/my-agent\"  # LiteLLM proxy + agent name\nheaders = {\"Authorization\": \"Bearer sk-1234\"}    # LiteLLM Virtual Key\n\nasync with httpx.AsyncClient(headers=headers) as httpx_client:\n    resolver = A2ACardResolver(httpx_client=httpx_client, base_url=base_url)\n    agent_card = await resolver.get_agent_card()\n    client = A2AClient(httpx_client=httpx_client, agent_card=agent_card)\n\n    request = SendMessageRequest(\n        id=str(uuid4()),\n        params=MessageSendParams(\n            message={\n                \"role\": \"user\",\n                \"parts\": [{\"kind\": \"text\", \"text\": \"Hello!\"}],\n                \"messageId\": uuid4().hex,\n            }\n        )\n    )\n    response = await client.send_message(request)\n```\n\n[**Docs: A2A Agent Gateway**](https:\/\/docs.litellm.ai\/docs\/a2a)\n\n<\/details>\n\n<details>\n<summary><b>MCP Tools<\/b> - Connect MCP servers to any LLM (Python SDK + AI Gateway)<\/summary>\n\n### Python SDK - MCP Bridge\n\n```python\nfrom mcp import ClientSession, StdioServerParameters\nfrom mcp.client.stdio import stdio_client\nfrom litellm import experimental_mcp_client\nimport litellm\n\nserver_params = StdioServerParameters(command=\"python\", args=[\"mcp_server.py\"])\n\nasync with stdio_client(server_params) as (read, write):\n    async with ClientSession(read, write) as session:\n        await session.initialize()\n\n        # Load MCP tools in OpenAI format\n        tools = await experimental_mcp_client.load_mcp_tools(session=session, format=\"openai\")\n\n        # Use with any LiteLLM model\n        response = await litellm.acompletion(\n            model=\"gpt-4o\",\n            messages=[{\"role\": \"user\", \"content\": \"What's 3 + 5?\"}],\n            tools=tools\n        )\n```\n\n### AI Gateway - MCP Gateway\n\n**Step 1.** [Add your MCP Server to the AI Gateway](https:\/\/docs.litellm.ai\/docs\/mcp#adding-your-mcp)\n\n**Step 2.** Call MCP tools via `\/chat\/completions`\n\n```bash\ncurl -X POST 'http:\/\/0.0.0.0:4000\/v1\/chat\/completions' \\\n  -H 'Authorization: Bearer sk-1234' \\\n  -H 'Content-Type: application\/json' \\\n  -d '{\n    \"model\": \"gpt-4o\",\n    \"messages\": [{\"role\": \"user\", \"content\": \"Summarize the latest open PR\"}],\n    \"tools\": [{\n      \"type\": \"mcp\",\n      \"server_url\": \"litellm_proxy\/mcp\/github\",\n      \"server_label\": \"github_mcp\",\n      \"require_approval\": \"never\"\n    }]\n  }'\n```\n\n### Use with Cursor IDE\n\n```json\n{\n  \"mcpServers\": {\n    \"LiteLLM\": {\n      \"url\": \"http:\/\/localhost:4000\/mcp\/\",\n      \"headers\": {\n        \"x-litellm-api-key\": \"Bearer sk-1234\"\n      }\n    }\n  }\n}\n```\n\n[**Docs: MCP Gateway**](https:\/\/docs.litellm.ai\/docs\/mcp)\n\n<\/details>\n\n---\n\n## How to use LiteLLM\n\nYou can use LiteLLM through either the Proxy Server or Python SDK. Both gives you a unified interface to access multiple LLMs (100+ LLMs). Choose the option that best fits your needs:\n\n<table style={{width: '100%', tableLayout: 'fixed'}}>\n<thead>\n<tr>\n<th style={{width: '14%'}}><\/th>\n<th style={{width: '43%'}}><strong><a href=\"https:\/\/docs.litellm.ai\/docs\/simple_proxy\">LiteLLM AI Gateway<\/a><\/strong><\/th>\n<th style={{width: '43%'}}><strong><a href=\"https:\/\/docs.litellm.ai\/docs\/\">LiteLLM Python SDK<\/a><\/strong><\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td style={{width: '14%'}}><strong>Use Case<\/strong><\/td>\n<td style={{width: '43%'}}>Central service (LLM Gateway) to access multiple LLMs<\/td>\n<td style={{width: '43%'}}>Use LiteLLM directly in your Python code<\/td>\n<\/tr>\n<tr>\n<td style={{width: '14%'}}><strong>Who Uses It?<\/strong><\/td>\n<td style={{width: '43%'}}>Gen AI Enablement \/ ML Platform Teams<\/td>\n<td style={{width: '43%'}}>Developers building LLM projects<\/td>\n<\/tr>\n<tr>\n<td style={{width: '14%'}}><strong>Key Features<\/strong><\/td>\n<td style={{width: '43%'}}>Centralized API gateway with authentication and authorization, multi-tenant cost tracking and spend management per project\/user, per-project customization (logging, guardrails, caching), virtual keys for secure access control, admin dashboard UI for monitoring and management<\/td>\n<td style={{width: '43%'}}>Direct Python library integration in your codebase, Router with retry\/fallback logic across multiple deployments (e.g. Azure\/OpenAI) - <a href=\"https:\/\/docs.litellm.ai\/docs\/routing\">Router<\/a>, application-level load balancing and cost tracking, exception handling with OpenAI-compatible errors, observability callbacks (Lunary, MLflow, Langfuse, etc.)<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n\nLiteLLM Performance: **8ms P95 latency** at 1k RPS (See benchmarks [here](https:\/\/docs.litellm.ai\/docs\/benchmarks))\n\n[**Jump to LiteLLM Proxy (LLM Gateway) Docs**](https:\/\/docs.litellm.ai\/docs\/simple_proxy) <br>\n[**Jump to Supported LLM Providers**](https:\/\/docs.litellm.ai\/docs\/providers)\n\n**Stable Release:** Use docker images with the `-stable` tag. These have undergone 12 hour load tests, before being published. [More information about the release cycle here](https:\/\/docs.litellm.ai\/docs\/proxy\/release_cycle)\n\nSupport for more providers. Missing a provider or LLM Platform, raise a [feature request](https:\/\/github.com\/BerriAI\/litellm\/issues\/new?assignees=&labels=enhancement&projects=&template=feature_request.yml&title=%5BFeature%5D%3A+).\n\n## OSS Adopters \n\n<table>\n  <tr>\n    <td><img height=\"60\" alt=\"Stripe\" src=\"https:\/\/github.com\/user-attachments\/assets\/f7296d4f-9fbd-460d-9d05-e4df31697c4b\" \/><\/td>\n    <td><img height=\"60\" alt=\"image\" src=\"https:\/\/github.com\/user-attachments\/assets\/436fca71-988b-40bb-b5fe-8450c80fdbd0\" \/><\/td>\n    <td><img height=\"60\" alt=\"Google ADK\" src=\"https:\/\/github.com\/user-attachments\/assets\/caf270a2-5aee-45c4-8222-41a2070c4f19\" \/><\/td>\n    <td><img height=\"60\" alt=\"Greptile\" src=\"https:\/\/github.com\/user-attachments\/assets\/0be4bd8a-7cfa-48d3-9090-f415fe948280\" \/><\/td>\n    <td><img height=\"60\" alt=\"OpenHands\" src=\"https:\/\/github.com\/user-attachments\/assets\/a6150c4c-149e-4cae-888b-8b92be6e003f\" \/><\/td>\n    <td><h2>Netflix<\/h2><\/td>\n    <td><img height=\"60\" alt=\"OpenAI Agents SDK\" src=\"https:\/\/github.com\/user-attachments\/assets\/c02f7be0-8c2e-4d27-aea7-7c024bfaebc0\" \/><\/td>\n  <\/tr>\n<\/table>\n\n## Supported Providers ([Website Supported Models](https:\/\/models.litellm.ai\/) | [Docs](https:\/\/docs.litellm.ai\/docs\/providers))\n\n| Provider                                                                            | `\/chat\/completions` | `\/messages` | `\/responses` | `\/embeddings` | `\/image\/generations` | `\/audio\/transcriptions` | `\/audio\/speech` | `\/moderations` | `\/batches` | `\/rerank` |\n|-------------------------------------------------------------------------------------|---------------------|-------------|--------------|---------------|----------------------|-------------------------|-----------------|----------------|-----------|-----------|\n| [Abliteration (`abliteration`)](https:\/\/docs.litellm.ai\/docs\/providers\/abliteration) | \u2705 |  |  |  |  |  |  |  |  |  |\n| [AI\/ML API (`aiml`)](https:\/\/docs.litellm.ai\/docs\/providers\/aiml) | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |\n| [AI21 (`ai21`)](https:\/\/docs.litellm.ai\/docs\/providers\/ai21) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [AI21 Chat (`ai21_chat`)](https:\/\/docs.litellm.ai\/docs\/providers\/ai21) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Aleph Alpha](https:\/\/docs.litellm.ai\/docs\/providers\/aleph_alpha) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Amazon Nova](https:\/\/docs.litellm.ai\/docs\/providers\/amazon_nova) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Anthropic (`anthropic`)](https:\/\/docs.litellm.ai\/docs\/providers\/anthropic) | \u2705 | \u2705 | \u2705 |  |  |  |  |  | \u2705 |  |\n| [Anthropic Text (`anthropic_text`)](https:\/\/docs.litellm.ai\/docs\/providers\/anthropic) | \u2705 | \u2705 | \u2705 |  |  |  |  |  | \u2705 |  |\n| [Anyscale](https:\/\/docs.litellm.ai\/docs\/providers\/anyscale) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [AssemblyAI (`assemblyai`)](https:\/\/docs.litellm.ai\/docs\/pass_through\/assembly_ai) | \u2705 | \u2705 | \u2705 |  |  | \u2705 |  |  |  |  |\n| [Auto Router (`auto_router`)](https:\/\/docs.litellm.ai\/docs\/proxy\/auto_routing) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [AWS - Bedrock (`bedrock`)](https:\/\/docs.litellm.ai\/docs\/providers\/bedrock) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  | \u2705 |\n| [AWS - Sagemaker (`sagemaker`)](https:\/\/docs.litellm.ai\/docs\/providers\/aws_sagemaker) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |\n| [Azure (`azure`)](https:\/\/docs.litellm.ai\/docs\/providers\/azure) | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 |  |\n| [Azure AI (`azure_ai`)](https:\/\/docs.litellm.ai\/docs\/providers\/azure_ai) | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 |  |\n| [Azure Text (`azure_text`)](https:\/\/docs.litellm.ai\/docs\/providers\/azure) | \u2705 | \u2705 | \u2705 |  |  | \u2705 | \u2705 | \u2705 | \u2705 |  |\n| [Baseten (`baseten`)](https:\/\/docs.litellm.ai\/docs\/providers\/baseten) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Bytez (`bytez`)](https:\/\/docs.litellm.ai\/docs\/providers\/bytez) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Cerebras (`cerebras`)](https:\/\/docs.litellm.ai\/docs\/providers\/cerebras) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Clarifai (`clarifai`)](https:\/\/docs.litellm.ai\/docs\/providers\/clarifai) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Cloudflare AI Workers (`cloudflare`)](https:\/\/docs.litellm.ai\/docs\/providers\/cloudflare_workers) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Codestral (`codestral`)](https:\/\/docs.litellm.ai\/docs\/providers\/codestral) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Cohere (`cohere`)](https:\/\/docs.litellm.ai\/docs\/providers\/cohere) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  | \u2705 |\n| [Cohere Chat (`cohere_chat`)](https:\/\/docs.litellm.ai\/docs\/providers\/cohere) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [CometAPI (`cometapi`)](https:\/\/docs.litellm.ai\/docs\/providers\/cometapi) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |\n| [CompactifAI (`compactifai`)](https:\/\/docs.litellm.ai\/docs\/providers\/compactifai) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Custom (`custom`)](https:\/\/docs.litellm.ai\/docs\/providers\/custom_llm_server) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Custom OpenAI (`custom_openai`)](https:\/\/docs.litellm.ai\/docs\/providers\/openai_compatible) | \u2705 | \u2705 | \u2705 |  |  | \u2705 | \u2705 | \u2705 | \u2705 |  |\n| [Dashscope (`dashscope`)](https:\/\/docs.litellm.ai\/docs\/providers\/dashscope) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Databricks (`databricks`)](https:\/\/docs.litellm.ai\/docs\/providers\/databricks) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [DataRobot (`datarobot`)](https:\/\/docs.litellm.ai\/docs\/providers\/datarobot) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Deepgram (`deepgram`)](https:\/\/docs.litellm.ai\/docs\/providers\/deepgram) | \u2705 | \u2705 | \u2705 |  |  | \u2705 |  |  |  |  |\n| [DeepInfra (`deepinfra`)](https:\/\/docs.litellm.ai\/docs\/providers\/deepinfra) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Deepseek (`deepseek`)](https:\/\/docs.litellm.ai\/docs\/providers\/deepseek) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [ElevenLabs (`elevenlabs`)](https:\/\/docs.litellm.ai\/docs\/providers\/elevenlabs) | \u2705 | \u2705 | \u2705 |  |  | \u2705 | \u2705 |  |  |  |\n| [Empower (`empower`)](https:\/\/docs.litellm.ai\/docs\/providers\/empower) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Fal AI (`fal_ai`)](https:\/\/docs.litellm.ai\/docs\/providers\/fal_ai) | \u2705 | \u2705 | \u2705 |  | \u2705 |  |  |  |  |  |\n| [Featherless AI (`featherless_ai`)](https:\/\/docs.litellm.ai\/docs\/providers\/featherless_ai) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Fireworks AI (`fireworks_ai`)](https:\/\/docs.litellm.ai\/docs\/providers\/fireworks_ai) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [FriendliAI (`friendliai`)](https:\/\/docs.litellm.ai\/docs\/providers\/friendliai) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Galadriel (`galadriel`)](https:\/\/docs.litellm.ai\/docs\/providers\/galadriel) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [GitHub Copilot (`github_copilot`)](https:\/\/docs.litellm.ai\/docs\/providers\/github_copilot) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |\n| [GitHub Models (`github`)](https:\/\/docs.litellm.ai\/docs\/providers\/github) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Google - PaLM](https:\/\/docs.litellm.ai\/docs\/providers\/palm) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Google - Vertex AI (`vertex_ai`)](https:\/\/docs.litellm.ai\/docs\/providers\/vertex) | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |\n| [Google AI Studio - Gemini (`gemini`)](https:\/\/docs.litellm.ai\/docs\/providers\/gemini) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [GradientAI (`gradient_ai`)](https:\/\/docs.litellm.ai\/docs\/providers\/gradient_ai) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Groq AI (`groq`)](https:\/\/docs.litellm.ai\/docs\/providers\/groq) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Heroku (`heroku`)](https:\/\/docs.litellm.ai\/docs\/providers\/heroku) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Hosted VLLM (`hosted_vllm`)](https:\/\/docs.litellm.ai\/docs\/providers\/vllm) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Huggingface (`huggingface`)](https:\/\/docs.litellm.ai\/docs\/providers\/huggingface) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  | \u2705 |\n| [Hyperbolic (`hyperbolic`)](https:\/\/docs.litellm.ai\/docs\/providers\/hyperbolic) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [IBM - Watsonx.ai (`watsonx`)](https:\/\/docs.litellm.ai\/docs\/providers\/watsonx) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |\n| [Infinity (`infinity`)](https:\/\/docs.litellm.ai\/docs\/providers\/infinity) |  |  |  | \u2705 |  |  |  |  |  |  |\n| [Jina AI (`jina_ai`)](https:\/\/docs.litellm.ai\/docs\/providers\/jina_ai) |  |  |  | \u2705 |  |  |  |  |  |  |\n| [Lambda AI (`lambda_ai`)](https:\/\/docs.litellm.ai\/docs\/providers\/lambda_ai) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Lemonade (`lemonade`)](https:\/\/docs.litellm.ai\/docs\/providers\/lemonade) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [LiteLLM Proxy (`litellm_proxy`)](https:\/\/docs.litellm.ai\/docs\/providers\/litellm_proxy) | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |\n| [Llamafile (`llamafile`)](https:\/\/docs.litellm.ai\/docs\/providers\/llamafile) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [LM Studio (`lm_studio`)](https:\/\/docs.litellm.ai\/docs\/providers\/lm_studio) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Maritalk (`maritalk`)](https:\/\/docs.litellm.ai\/docs\/providers\/maritalk) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Meta - Llama API (`meta_llama`)](https:\/\/docs.litellm.ai\/docs\/providers\/meta_llama) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Mistral AI API (`mistral`)](https:\/\/docs.litellm.ai\/docs\/providers\/mistral) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |\n| [Moonshot (`moonshot`)](https:\/\/docs.litellm.ai\/docs\/providers\/moonshot) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Morph (`morph`)](https:\/\/docs.litellm.ai\/docs\/providers\/morph) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Nebius AI Studio (`nebius`)](https:\/\/docs.litellm.ai\/docs\/providers\/nebius) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |\n| [NLP Cloud (`nlp_cloud`)](https:\/\/docs.litellm.ai\/docs\/providers\/nlp_cloud) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Novita AI (`novita`)](https:\/\/novita.ai\/models\/llm?utm_source=github_litellm&utm_medium=github_readme&utm_campaign=github_link) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Nscale (`nscale`)](https:\/\/docs.litellm.ai\/docs\/providers\/nscale) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Nvidia NIM (`nvidia_nim`)](https:\/\/docs.litellm.ai\/docs\/providers\/nvidia_nim) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [OCI (`oci`)](https:\/\/docs.litellm.ai\/docs\/providers\/oci) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Ollama (`ollama`)](https:\/\/docs.litellm.ai\/docs\/providers\/ollama) | \u2705 | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |\n| [Ollama Chat (`ollama_chat`)](https:\/\/docs.litellm.ai\/docs\/providers\/ollama) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Oobabooga (`oobabooga`)](https:\/\/docs.litellm.ai\/docs\/providers\/openai_compatible) | \u2705 | \u2705 | \u2705 |  |  | \u2705 | \u2705 | \u2705 | \u2705 |  |\n| [OpenAI (`openai`)](https:\/\/docs.litellm.ai\/docs\/providers\/openai) | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 | \u2705 |  |\n| [OpenAI-like (`openai_like`)](https:\/\/docs.litellm.ai\/docs\/providers\/openai_compatible) |  |  |  | \u2705 |  |  |  |  |  |  |\n| [OpenRouter (`openrouter`)](https:\/\/docs.litellm.ai\/docs\/providers\/openrouter) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [OVHCloud AI Endpoints (`ovhcloud`)](https:\/\/docs.litellm.ai\/docs\/providers\/ovhcloud) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Perplexity AI (`perplexity`)](https:\/\/docs.litellm.ai\/docs\/providers\/perplexity) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Petals (`petals`)](https:\/\/docs.litellm.ai\/docs\/providers\/petals) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Predibase (`predibase`)](https:\/\/docs.litellm.ai\/docs\/providers\/predibase) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Recraft (`recraft`)](https:\/\/docs.litellm.ai\/docs\/providers\/recraft) |  |  |  |  | \u2705 |  |  |  |  |  |\n| [Replicate (`replicate`)](https:\/\/docs.litellm.ai\/docs\/providers\/replicate) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Sagemaker Chat (`sagemaker_chat`)](https:\/\/docs.litellm.ai\/docs\/providers\/aws_sagemaker) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Sambanova (`sambanova`)](https:\/\/docs.litellm.ai\/docs\/providers\/sambanova) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Snowflake (`snowflake`)](https:\/\/docs.litellm.ai\/docs\/providers\/snowflake) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Text Completion Codestral (`text-completion-codestral`)](https:\/\/docs.litellm.ai\/docs\/providers\/codestral) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Text Completion OpenAI (`text-completion-openai`)](https:\/\/docs.litellm.ai\/docs\/providers\/text_completion_openai) | \u2705 | \u2705 | \u2705 |  |  | \u2705 | \u2705 | \u2705 | \u2705 |  |\n| [Together AI (`together_ai`)](https:\/\/docs.litellm.ai\/docs\/providers\/togetherai) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Topaz (`topaz`)](https:\/\/docs.litellm.ai\/docs\/providers\/topaz) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Triton (`triton`)](https:\/\/docs.litellm.ai\/docs\/providers\/triton-inference-server) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [V0 (`v0`)](https:\/\/docs.litellm.ai\/docs\/providers\/v0) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Vercel AI Gateway (`vercel_ai_gateway`)](https:\/\/docs.litellm.ai\/docs\/providers\/vercel_ai_gateway) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [VLLM (`vllm`)](https:\/\/docs.litellm.ai\/docs\/providers\/vllm) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Volcengine (`volcengine`)](https:\/\/docs.litellm.ai\/docs\/providers\/volcano) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Voyage AI (`voyage`)](https:\/\/docs.litellm.ai\/docs\/providers\/voyage) |  |  |  | \u2705 |  |  |  |  |  |  |\n| [WandB Inference (`wandb`)](https:\/\/docs.litellm.ai\/docs\/providers\/wandb_inference) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Watsonx Text (`watsonx_text`)](https:\/\/docs.litellm.ai\/docs\/providers\/watsonx) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [xAI (`xai`)](https:\/\/docs.litellm.ai\/docs\/providers\/xai) | \u2705 | \u2705 | \u2705 |  |  |  |  |  |  |  |\n| [Xinference (`xinference`)](https:\/\/docs.litellm.ai\/docs\/providers\/xinference) |  |  |  | \u2705 |  |  |  |  |  |  |\n\n[**Read the Docs**](https:\/\/docs.litellm.ai\/docs\/)\n\n## Run in Developer mode\n### Services\n1. Setup .env file in root\n2. Run dependant services `docker-compose up db prometheus`\n\n### Backend\n1. (In root) create virtual environment `python -m venv .venv`\n2. Activate virtual environment `source .venv\/bin\/activate`\n3. Install dependencies `pip install -e \".[all]\"`\n4. `pip install prisma`\n5. `prisma generate`\n6. Start proxy backend `python litellm\/proxy\/proxy_cli.py`\n\n### Frontend\n1. Navigate to `ui\/litellm-dashboard`\n2. Install dependencies `npm install`\n3. Run `npm run dev` to start the dashboard\n\n# Enterprise\nFor companies that need better security, user management and professional support\n\n[Talk to founders](https:\/\/enterprise.litellm.ai\/demo)\n\nThis covers:\n- \u2705 **Features under the [LiteLLM Commercial License](https:\/\/docs.litellm.ai\/docs\/proxy\/enterprise):**\n- \u2705 **Feature Prioritization**\n- \u2705 **Custom Integrations**\n- \u2705 **Professional Support - Dedicated discord + slack**\n- \u2705 **Custom SLAs**\n- \u2705 **Secure access with Single Sign-On**\n\n# Contributing\n\nWe welcome contributions to LiteLLM! Whether you're fixing bugs, adding features, or improving documentation, we appreciate your help.\n\n## Quick Start for Contributors\n\nThis requires poetry to be installed.\n\n```bash\ngit clone https:\/\/github.com\/BerriAI\/litellm.git\ncd litellm\nmake install-dev    # Install development dependencies\nmake format         # Format your code\nmake lint           # Run all linting checks\nmake test-unit      # Run unit tests\nmake format-check   # Check formatting only\n```\n\nFor detailed contributing guidelines, see [CONTRIBUTING.md](CONTRIBUTING.md).\n\n## Code Quality \/ Linting\n\nLiteLLM follows the [Google Python Style Guide](https:\/\/google.github.io\/styleguide\/pyguide.html).\n\nOur automated checks include:\n- **Black** for code formatting\n- **Ruff** for linting and code quality\n- **MyPy** for type checking\n- **Circular import detection**\n- **Import safety checks**\n\n\nAll these checks must pass before your PR can be merged.\n\n\n# Support \/ talk with founders\n\n- [Schedule Demo \ud83d\udc4b](https:\/\/calendly.com\/d\/4mp-gd3-k5k\/berriai-1-1-onboarding-litellm-hosted-version)\n- [Community Discord \ud83d\udcad](https:\/\/discord.gg\/wuPM9dRgDw)\n- [Community Slack \ud83d\udcad](https:\/\/www.litellm.ai\/support)\n- Our emails \u2709\ufe0f ishaan@berri.ai \/ krrish@berri.ai\n\n# Why did we build this\n\n- **Need for simplicity**: Our code started to get extremely complicated managing & translating calls between Azure, OpenAI and Cohere.\n\n# Contributors\n\n<!-- ALL-CONTRIBUTORS-LIST:START - Do not remove or modify this section -->\n<!-- prettier-ignore-start -->\n<!-- markdownlint-disable -->\n\n<!-- markdownlint-restore -->\n<!-- prettier-ignore-end -->\n\n<!-- ALL-CONTRIBUTORS-LIST:END -->\n\n<a href=\"https:\/\/github.com\/BerriAI\/litellm\/graphs\/contributors\">\n  <img src=\"https:\/\/contrib.rocks\/image?repo=BerriAI\/litellm\" \/>\n<\/a>\n\n\n","description_content_type":"text\/markdown","docs_url":null,"download_url":null,"downloads":{"last_day":-1,"last_month":-1,"last_week":-1},"dynamic":null,"home_page":null,"keywords":null,"license":"MIT","license_expression":null,"license_files":["LICENSE"],"maintainer":null,"maintainer_email":null,"name":"litellm","package_url":"https:\/\/pypi.org\/project\/litellm\/","platform":null,"project_url":"https:\/\/pypi.org\/project\/litellm\/","project_urls":{"Documentation":"https:\/\/docs.litellm.ai","Homepage":"https:\/\/litellm.ai","Repository":"https:\/\/github.com\/BerriAI\/litellm","documentation":"https:\/\/docs.litellm.ai","repository":"https:\/\/github.com\/BerriAI\/litellm"},"provides_extra":["caching","extra-proxy","google","grpc","mlflow","proxy","semantic-router","utils"],"release_url":"https:\/\/pypi.org\/project\/litellm\/1.83.0\/","requires_dist":["PyJWT<3.0.0,>=2.12.0; python_version >= \"3.9\" and extra == \"proxy\"","a2a-sdk<0.4.0,>=0.3.22; python_version >= \"3.10\" and extra == \"extra-proxy\"","aiohttp>=3.10","apscheduler<4.0.0,>=3.10.4; extra == \"proxy\"","azure-identity<2.0.0,>=1.15.0; python_version >= \"3.9\" and (extra == \"proxy\" or extra == \"extra-proxy\")","azure-keyvault-secrets<5.0.0,>=4.8.0; extra == \"extra-proxy\"","azure-storage-blob<13.0.0,>=12.25.1; extra == \"proxy\"","backoff; extra == \"proxy\"","boto3<2.0.0,>=1.40.76; extra == \"proxy\"","click","cryptography; extra == \"proxy\"","diskcache<6.0.0,>=5.6.1; extra == \"caching\"","fastapi>=0.120.1; extra == \"proxy\"","fastapi-sso<0.17.0,>=0.16.0; extra == \"proxy\"","fastuuid>=0.13.0","google-cloud-aiplatform>=1.38.0; extra == \"google\"","google-cloud-iam<3.0.0,>=2.19.1; extra == \"extra-proxy\"","google-cloud-kms<3.0.0,>=2.21.3; extra == \"extra-proxy\"","grpcio!=1.68.*,!=1.69.*,!=1.70.*,!=1.71.0,!=1.71.1,!=1.72.0,!=1.72.1,!=1.73.0,>=1.62.3; python_version < \"3.14\" and extra == \"grpc\"","grpcio>=1.75.0; python_version >= \"3.14\" and extra == \"grpc\"","gunicorn<24.0.0,>=23.0.0; extra == \"proxy\"","httpx>=0.23.0","importlib-metadata>=6.8.0","jinja2<4.0.0,>=3.1.2","jsonschema<5.0.0,>=4.23.0","litellm-enterprise==0.1.35; extra == \"proxy\"","litellm-proxy-extras<0.5.0,>=0.4.62; extra == \"proxy\"","mcp<2.0.0,>=1.25.0; python_version >= \"3.10\" and extra == \"proxy\"","mlflow>3.1.4; python_version >= \"3.10\" and extra == \"mlflow\"","numpydoc; extra == \"utils\"","openai>=2.8.0","orjson<4.0.0,>=3.9.7; extra == \"proxy\"","polars<2.0.0,>=1.31.0; python_version >= \"3.10\" and extra == \"proxy\"","prisma<0.12.0,>=0.11.0; extra == \"extra-proxy\"","pydantic<3.0.0,>=2.5.0","pynacl<2.0.0,>=1.5.0; extra == \"proxy\"","pyroscope-io<0.9,>=0.8; sys_platform != \"win32\" and extra == \"proxy\"","python-dotenv>=0.2.0","python-multipart>=0.0.20; extra == \"proxy\"","pyyaml<7.0.0,>=6.0.1; extra == \"proxy\"","redisvl<0.5.0,>=0.4.1; (python_version >= \"3.9\" and python_version < \"3.14\") and extra == \"extra-proxy\"","resend>=0.8.0; extra == \"extra-proxy\"","rich<14.0.0,>=13.7.1; extra == \"proxy\"","rq; extra == \"proxy\"","semantic-router>=0.1.12; (python_version >= \"3.9\" and python_version < \"3.14\") and extra == \"semantic-router\"","soundfile<0.13.0,>=0.12.1; extra == \"proxy\"","tiktoken>=0.7.0","tokenizers","uvicorn<1.0.0,>=0.32.1; extra == \"proxy\"","uvloop<0.22.0,>=0.21.0; sys_platform != \"win32\" and extra == \"proxy\"","websockets<16.0.0,>=15.0.1; extra == \"proxy\""],"requires_python":"<4.0,>=3.9","summary":"Library to easily interface with LLM API providers","version":"1.83.0","yanked":false,"yanked_reason":null},"last_serial":43868637,"ownership":{"organization":"LiteLLM","roles":[]},"urls":[{"comment_text":null,"core-metadata":{"sha256":"e33bfb7a17202dc1e17aa995e3204b4856d903e5bd7a10b6fbd747fbe6140ac4"},"digests":{"blake2b_256":"192ca670cc050fcd6f45c6199eb99e259c73aea92edba8d5c2fc1b3686d36217","md5":"6f01c1e0bc231ecdb937eb4388f38232","sha256":"88c536d339248f3987571493015784671ba3f193a328e1ea6780dbebaa2094a8"},"downloads":-1,"filename":"litellm-1.83.0-py3-none-any.whl","has_sig":false,"md5_digest":"6f01c1e0bc231ecdb937eb4388f38232","packagetype":"bdist_wheel","python_version":"py3","requires_python":"<4.0,>=3.9","size":15610306,"upload_time":"2026-03-31T05:08:21","upload_time_iso_8601":"2026-03-31T05:08:21.987889Z","url":"https:\/\/files.pythonhosted.org\/packages\/19\/2c\/a670cc050fcd6f45c6199eb99e259c73aea92edba8d5c2fc1b3686d36217\/litellm-1.83.0-py3-none-any.whl","yanked":false,"yanked_reason":null},{"comment_text":null,"core-metadata":false,"digests":{"blake2b_256":"22926ce9737554994ca8e536e5f4f6a87cc7c4774b656c9eb9add071caf7d54b","md5":"093c5c08fb9ddcd3b17d67518ba66d2b","sha256":"860bebc76c4bb27b4cf90b4a77acd66dba25aced37e3db98750de8a1766bfb7a"},"downloads":-1,"filename":"litellm-1.83.0.tar.gz","has_sig":false,"md5_digest":"093c5c08fb9ddcd3b17d67518ba66d2b","packagetype":"sdist","python_version":"source","requires_python":"<4.0,>=3.9","size":17333062,"upload_time":"2026-03-31T05:08:25","upload_time_iso_8601":"2026-03-31T05:08:25.331449Z","url":"https:\/\/files.pythonhosted.org\/packages\/22\/92\/6ce9737554994ca8e536e5f4f6a87cc7c4774b656c9eb9add071caf7d54b\/litellm-1.83.0.tar.gz","yanked":false,"yanked_reason":null}],"vulnerabilities":[{"aliases":["CVE-2026-12773"],"details":"A weakness has been identified in BerriAI litellm up to 1.59.8. Affected is the function UserAPIKeyAuth of the file litellm\/proxy\/_experimental\/mcp_server\/auth\/user_api_key_auth_mcp.py of the component MCP Proxy. Executing a manipulation can lead to improper authentication. The attack may be launched remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure.","fixed_in":["1.84.0"],"id":"GHSA-4jcj-7x88-m979","link":"https:\/\/osv.dev\/vulnerability\/GHSA-4jcj-7x88-m979","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-42203","PYSEC-2026-2602"],"details":"### Impact\nThe `POST \/prompts\/test` endpoint accepted user-supplied prompt templates and rendered them without sandboxing. A crafted template could run arbitrary code inside the LiteLLM Proxy process.\n\nThe endpoint only checks that the caller presents a valid proxy API key, so any authenticated user could reach it. Depending on how the proxy is deployed, this could expose secrets in the process environment (such as provider API keys or database credentials) and allow commands to be run on the host.\n\nProxy deployments running an affected version are in scope.\n\n### Patches\nThe issue is fixed in **`1.83.7-stable`**. The fix switches the prompt template renderer to a sandboxed environment that blocks the attributes this attack relies on.\n\nLiteLLM recommends upgrading to `1.83.7-stable` or later.\n\n### Workarounds\nIf upgrading is not immediately possible:\n\n1. Block `POST \/prompts\/test` at your reverse proxy or API gateway.\n2. Review and rotate API keys that should not have access to prompt management routes.","fixed_in":["1.83.7"],"id":"GHSA-xqmj-j6mv-4862","link":"https:\/\/osv.dev\/vulnerability\/GHSA-xqmj-j6mv-4862","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-42208","PYSEC-2026-391"],"details":"### Impact\n\nA database query used during proxy API key checks mixed the caller-supplied key value into the query text instead of passing it as a separate parameter. An unauthenticated attacker could send a specially crafted `Authorization` header to any LLM API route (for example `POST \/chat\/completions`) and reach this query through the proxy's error-handling path.\n\nAn attacker could read data from the proxy's database and may be able to modify it, leading to unauthorised access to the proxy and the credentials it manages.\n\n### Patches\n\nFixed in **`1.83.7`**. The caller-supplied value is now always passed to the database as a separate parameter. Upgrade to `1.83.7` or later.\n\n### Workarounds\n\nIf upgrading is not immediately possible, set `disable_error_logs: true` under `general_settings`. This removes the path through which unauthenticated input reaches the vulnerable query.\n\n### References\n\n- Patched release: [`v1.83.7-stable`](https:\/\/github.com\/BerriAI\/litellm\/releases\/tag\/v1.83.7-stable)\n\n**Discovery Credit**: Tencent YunDing Security Lab","fixed_in":["1.83.7"],"id":"GHSA-r75f-5x8p-qvmc","link":"https:\/\/osv.dev\/vulnerability\/GHSA-r75f-5x8p-qvmc","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-42271","PYSEC-2026-2599"],"details":"### Impact\n\nTwo endpoints used to preview an MCP server before saving it \u2014 `POST \/mcp-rest\/test\/connection` and `POST \/mcp-rest\/test\/tools\/list` \u2014 accepted a full server configuration in the request body, including the `command`, `args`, and `env` fields used by the stdio transport. When called with a stdio configuration, the endpoints attempted to connect, which spawned the supplied command as a subprocess on the proxy host with the privileges of the proxy process.\n\nThe endpoints were gated only by a valid proxy API key, with no role check. Any authenticated user \u2014 including holders of low-privilege internal-user keys \u2014 could therefore run arbitrary commands on the host.\n\n### Patches\n\nFixed in **`1.83.7`**. Both test endpoints now require the `PROXY_ADMIN` role, bringing them into line with the save endpoint.\n\n### Workarounds\n\nIf upgrading is not immediately possible, developers should block `POST \/mcp-rest\/test\/connection` and `POST \/mcp-rest\/test\/tools\/list` at their reverse proxy or API gateway.","fixed_in":["1.83.7"],"id":"GHSA-v4p8-mg3p-g94g","link":"https:\/\/osv.dev\/vulnerability\/GHSA-v4p8-mg3p-g94g","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-40217","PYSEC-2026-2601"],"details":"### Impact\n\nThe `POST \/guardrails\/test_custom_code` endpoint runs user-supplied Python inside a hand-rolled sandbox. The sandbox can be escaped using bytecode-level techniques, allowing arbitrary code execution in the proxy process \u2014 which runs as root in the default Docker image.\n\n**Reaching the endpoint requires a proxy-admin credential** in default configurations.\n\n### Patches\n\nFixed in **`1.83.11`**. The hand-rolled sandbox has been replaced with `RestrictedPython`. Upgrade to `1.83.11` or later.\n\n### Workarounds\n\nIf upgrading is not immediately possible, block `POST \/guardrails\/test_custom_code` at your reverse proxy or API gateway.\n\n### References\n\n- Patched release: [`v1.83.10-stable`](https:\/\/github.com\/BerriAI\/litellm\/releases\/tag\/v1.83.10-stable)","fixed_in":["1.83.10"],"id":"GHSA-wxxx-gvqv-xp7p","link":"https:\/\/osv.dev\/vulnerability\/GHSA-wxxx-gvqv-xp7p","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-49468","PYSEC-2026-388"],"details":"### Impact\n\nA Host-header parsing flaw in the LiteLLM proxy could, under specific conditions, allow unauthenticated access to protected management routes.\n\nThe auth layer derived the effective route from `request.url.path` in `litellm\/proxy\/auth\/auth_utils.py::get_request_route()`, which Starlette reconstructs from the `Host` header. A crafted `Host` could therefore make the auth gate evaluate a different route from the one FastAPI dispatched.\n\n**Most deployments are not affected.** The bypass is blocked by any upstream layer that validates or normalizes `Host`, such as:\n\n- a CDN or WAF, such as Cloudflare\n- a reverse proxy with `server_name` allowlists\n- a host-based load balancer\n\n**LiteLLM Cloud customers are not affected.**\n\n### Patches\n\nFixed in **`1.84.0`**. Upgrade to `1.84.0` or later. No configuration change is required.\n\n### Workarounds\n\nIf upgrading is not immediately possible, place the proxy behind an upstream component that validates or normalizes the `Host` header before forwarding (a CDN\/WAF, a reverse proxy with explicit `server_name` allowlists, or a cloud load balancer with host-based routing rules), or otherwise restrict network access to the proxy listener.\n\n### References\n\n- Patched release: [`v1.84.0`](https:\/\/github.com\/BerriAI\/litellm\/releases\/tag\/v1.84.0)\n\n**Discovery Credit**: Le The Thang (KCSC) and Kim Ngoc Chung (One Mount Group)","fixed_in":["1.84.0"],"id":"GHSA-4xpc-pv4p-pm3w","link":"https:\/\/osv.dev\/vulnerability\/GHSA-4xpc-pv4p-pm3w","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-47102","PYSEC-2026-2600"],"details":"LiteLLM prior to 1.83.10 allows a user to modify their own user_role via the \/user\/update endpoint. While the endpoint correctly restricts users to updating only their own account, it does not restrict which fields may be changed. A user who can reach this endpoint can set their role to proxy_admin, gaining full administrative access to LiteLLM including all users, teams, keys, models, and prompt history. Users with the org_admin role have legitimate access to this endpoint and can exploit this vulnerability without chaining any additional flaw.","fixed_in":["1.83.10"],"id":"GHSA-wpfp-gwwc-vwq6","link":"https:\/\/osv.dev\/vulnerability\/GHSA-wpfp-gwwc-vwq6","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-47101","PYSEC-2026-2598"],"details":"LiteLLM prior to 1.83.14 allows an authenticated internal_user to create API keys with access to routes that their role does not permit. When generating a key, the allowed_routes field is stored without verifying that the specified routes fall within the user's own permissions. A key created with access to admin-only routes can then be used to reach those routes successfully, bypassing the role-based access controls that would otherwise block the request, enabling full privilege escalation from internal_user to proxy_admin.","fixed_in":["1.83.14"],"id":"GHSA-qrc4-49gv-mv9m","link":"https:\/\/osv.dev\/vulnerability\/GHSA-qrc4-49gv-mv9m","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-42208","GHSA-r75f-5x8p-qvmc"],"details":"### Impact\n\nA database query used during proxy API key checks mixed the caller-supplied key value into the query text instead of passing it as a separate parameter. An unauthenticated attacker could send a specially crafted `Authorization` header to any LLM API route (for example `POST \/chat\/completions`) and reach this query through the proxy's error-handling path.\n\nAn attacker could read data from the proxy's database and may be able to modify it, leading to unauthorised access to the proxy and the credentials it manages.\n\n### Patches\n\nFixed in **`1.83.7`**. The caller-supplied value is now always passed to the database as a separate parameter. Upgrade to `1.83.7` or later.\n\n### Workarounds\n\nIf upgrading is not immediately possible, set `disable_error_logs: true` under `general_settings`. This removes the path through which unauthenticated input reaches the vulnerable query.\n\n### References\n\n- Patched release: [`v1.83.7-stable`](https:\/\/github.com\/BerriAI\/litellm\/releases\/tag\/v1.83.7-stable)\n \n**Discovery Credit**: Tencent YunDing Security Lab","fixed_in":["1.83.7"],"id":"PYSEC-2026-391","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-391","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-49468","GHSA-4xpc-pv4p-pm3w"],"details":"### Impact\n\nA Host-header parsing flaw in the LiteLLM proxy could, under specific conditions, allow unauthenticated access to protected management routes.\n \nThe auth layer derived the effective route from `request.url.path` in `litellm\/proxy\/auth\/auth_utils.py::get_request_route()`, which Starlette reconstructs from the `Host` header. A crafted `Host` could therefore make the auth gate evaluate a different route from the one FastAPI dispatched.\n \n**Most deployments are not affected.** The bypass is blocked by any upstream layer that validates or normalizes `Host`, such as:\n\n- a CDN or WAF, such as Cloudflare\n - a reverse proxy with `server_name` allowlists\n- a host-based load balancer\n\n **LiteLLM Cloud customers are not affected.**\n\n### Patches\n\nFixed in **`1.84.0`**. Upgrade to `1.84.0` or later. No configuration change is required.\n\n### Workarounds\n \nIf upgrading is not immediately possible, place the proxy behind an upstream component that validates or normalizes the `Host` header before forwarding (a CDN\/WAF, a reverse proxy with explicit `server_name` allowlists, or a cloud load balancer with host-based routing rules), or otherwise restrict network access to the proxy listener.\n\n ### References\n\n- Patched release: [`v1.84.0`](https:\/\/github.com\/BerriAI\/litellm\/releases\/tag\/v1.84.0)\n \n**Discovery Credit**: Le The Thang (KCSC) and Kim Ngoc Chung (One Mount Group)","fixed_in":["1.84.0"],"id":"PYSEC-2026-388","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-388","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-59823"],"details":"### Summary\nA server-side request forgery in LiteLLM Proxy lets an authenticated caller redirect the\nproxy's outbound request to a host of their choosing by smuggling an `api_base` inside the\n`user_config` request body, bypassing the existing parameter guard.\n\n### Details\nLiteLLM Proxy validates request bodies with `is_request_body_safe`, which blocks the\n`api_base` and `base_url` parameters but does not cover `user_config`. The `user_config`\nobject is used to build the outbound router for a request, so a caller can place an\n`api_base` inside it and reach an arbitrary host. The guard only inspected the two\ntop-level keys, so the same `api_base` nested inside `user_config` was never checked.\n\nExploitation requires a valid virtual key.\n\n### Impact\nAn authenticated caller can make the proxy issue server-side requests to internal or\nexternal hosts of their choosing, reaching endpoints the caller cannot otherwise access.\n\n### Affected \/ Patched\nAffected: `<= 1.83.8`\nPatched: `1.83.9`\n\n### Remediation\nUpgrade to 1.83.9 or later (released 2026-04-17).","fixed_in":["1.83.9"],"id":"GHSA-hx8v-g79f-8w5f","link":"https:\/\/osv.dev\/vulnerability\/GHSA-hx8v-g79f-8w5f","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-42203","GHSA-xqmj-j6mv-4862"],"details":"### Impact\nThe `POST \/prompts\/test` endpoint accepted user-supplied prompt templates and rendered them without sandboxing. A crafted template could run arbitrary code inside the LiteLLM Proxy process.\n\nThe endpoint only checks that the caller presents a valid proxy API key, so any authenticated user could reach it. Depending on how the proxy is deployed, this could expose secrets in the process environment (such as provider API keys or database credentials) and allow commands to be run on the host.\n\nProxy deployments running an affected version are in scope.\n\n### Patches\nThe issue is fixed in **`1.83.7-stable`**. The fix switches the prompt template renderer to a sandboxed environment that blocks the attributes this attack relies on.\n\nLiteLLM recommends upgrading to `1.83.7-stable` or later.\n\n### Workarounds\nIf upgrading is not immediately possible:\n\n1. Block `POST \/prompts\/test` at your reverse proxy or API gateway.\n2. Review and rotate API keys that should not have access to prompt management routes.","fixed_in":["1.83.7"],"id":"PYSEC-2026-2602","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-2602","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-40217","GHSA-wxxx-gvqv-xp7p"],"details":"### Impact\n\nThe `POST \/guardrails\/test_custom_code` endpoint runs user-supplied Python inside a hand-rolled sandbox. The sandbox can be escaped using bytecode-level techniques, allowing arbitrary code execution in the proxy process \u2014 which runs as root in the default Docker image.\n\n**Reaching the endpoint requires a proxy-admin credential** in default configurations.\n\n### Patches\n\nFixed in **`1.83.11`**. The hand-rolled sandbox has been replaced with `RestrictedPython`. Upgrade to `1.83.11` or later.\n\n### Workarounds\n\nIf upgrading is not immediately possible, block `POST \/guardrails\/test_custom_code` at your reverse proxy or API gateway.\n\n### References\n\n- Patched release: [`v1.83.10-stable`](https:\/\/github.com\/BerriAI\/litellm\/releases\/tag\/v1.83.10-stable)","fixed_in":["1.83.10"],"id":"PYSEC-2026-2601","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-2601","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-42271","GHSA-v4p8-mg3p-g94g"],"details":"### Impact\n\nTwo endpoints used to preview an MCP server before saving it \u2014 `POST \/mcp-rest\/test\/connection` and `POST \/mcp-rest\/test\/tools\/list` \u2014 accepted a full server configuration in the request body, including the `command`, `args`, and `env` fields used by the stdio transport. When called with a stdio configuration, the endpoints attempted to connect, which spawned the supplied command as a subprocess on the proxy host with the privileges of the proxy process.\n\nThe endpoints were gated only by a valid proxy API key, with no role check. Any authenticated user \u2014 including holders of low-privilege internal-user keys \u2014 could therefore run arbitrary commands on the host.\n\n### Patches\n\nFixed in **`1.83.7`**. Both test endpoints now require the `PROXY_ADMIN` role, bringing them into line with the save endpoint.\n\n### Workarounds\n\nIf upgrading is not immediately possible, developers should block `POST \/mcp-rest\/test\/connection` and `POST \/mcp-rest\/test\/tools\/list` at their reverse proxy or API gateway.","fixed_in":["1.83.7"],"id":"PYSEC-2026-2599","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-2599","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-47101","GHSA-qrc4-49gv-mv9m"],"details":"LiteLLM prior to 1.83.14 allows an authenticated internal_user to create API keys with access to routes that their role does not permit. When generating a key, the allowed_routes field is stored without verifying that the specified routes fall within the user's own permissions. A key created with access to admin-only routes can then be used to reach those routes successfully, bypassing the role-based access controls that would otherwise block the request, enabling full privilege escalation from internal_user to proxy_admin.","fixed_in":["1.83.14"],"id":"PYSEC-2026-2598","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-2598","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-47102","GHSA-wpfp-gwwc-vwq6"],"details":"LiteLLM prior to 1.83.10 allows a user to modify their own user_role via the \/user\/update endpoint. While the endpoint correctly restricts users to updating only their own account, it does not restrict which fields may be changed. A user who can reach this endpoint can set their role to proxy_admin, gaining full administrative access to LiteLLM including all users, teams, keys, models, and prompt history. Users with the org_admin role have legitimate access to this endpoint and can exploit this vulnerability without chaining any additional flaw.","fixed_in":["1.83.10"],"id":"PYSEC-2026-2600","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-2600","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-59820","PYSEC-2026-3477"],"details":"### Impact\n\nLiteLLM Skills archive extraction did not sufficiently validate file paths from uploaded skill ZIP archives. An authenticated user with access to LiteLLM LLM API routes, or a key whose `allowed_routes` includes `\/v1\/skills`, `anthropic_routes`, or `llm_api_routes`, could upload a crafted skill archive containing path traversal entries.\n\nWhen the skill was processed for execution, those entries could be written outside the intended extraction\/staging directory. This could allow arbitrary file write and may lead to code execution depending on deployment configuration and writable paths.\n\n### Patches\n\nThe issue is fixed in `1.83.7-stable`.\n\nLiteLLM recommens upgrading to `1.83.7-stable` or later.\n\n### Workarounds\n\nIf upgrading is not immediately possible:\n\n1. Block `POST \/v1\/skills` at your reverse proxy or API gateway.\n2. Restrict Skills API access to trusted users only.","fixed_in":["1.83.7"],"id":"GHSA-5jmr-gcrj-2c9q","link":"https:\/\/osv.dev\/vulnerability\/GHSA-5jmr-gcrj-2c9q","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-59822","PYSEC-2026-3479"],"details":"### Impact\n\nLiteLLM's MCP Streamable HTTP endpoint could allow an unauthenticated attacker to establish an authenticated MCP session using an arbitrary Bearer token.\n\nThe MCP auth handler supported OAuth2 passthrough for upstream MCP servers, but the fallback path could replace failed LiteLLM key validation with an empty `UserAPIKeyAuth()` object. This allowed requests with a fabricated `Authorization` header to reach MCP tooling without a valid LiteLLM key.\n\nAn attacker could use this to list and call configured MCP tools and access connected services exposed through MCP.\n\n### Patches\n\nThe issue is fixed in `1.84.0`.\n\nWe recommend upgrading to `1.84.0` or later.\n\n### Workarounds\n\nIf upgrading is not immediately possible, disable MCP routes or block access to `\/mcp\/` and related MCP endpoints at your reverse proxy or API gateway.\n\n### References\n\n* [v1.84.0](https:\/\/github.com\/BerriAI\/litellm\/releases\/tag\/v1.84.0)","fixed_in":["1.84.0"],"id":"GHSA-7488-6r32-c95q","link":"https:\/\/osv.dev\/vulnerability\/GHSA-7488-6r32-c95q","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-59819","PYSEC-2026-3476"],"details":"### Impact\n\nLiteLLM's `\/health\/test_connection` endpoint resolved request-supplied environment and OIDC file references in `litellm_params`. A proxy administrator, or another privileged caller with permission to test model connections, could cause LiteLLM to read files from the local filesystem via an `oidc\/file\/` reference.\n\nBecause exploitation requires privileged proxy access, this is treated as a defense-in-depth issue rather than a cross-tenant privilege bypass.\n\n### Patches\n\nThe issue is fixed in `1.83.10-stable`.\n\nLiteLLM recommend upgrading to `1.83.10-stable` or later.\n\n### Workarounds\n\nRestrict `\/health\/test_connection` access to trusted administrators only.","fixed_in":["1.83.10"],"id":"GHSA-4g5m-c9r5-49xf","link":"https:\/\/osv.dev\/vulnerability\/GHSA-4g5m-c9r5-49xf","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-59820","GHSA-5jmr-gcrj-2c9q"],"details":"### Impact\n\nLiteLLM Skills archive extraction did not sufficiently validate file paths from uploaded skill ZIP archives. An authenticated user with access to LiteLLM LLM API routes, or a key whose `allowed_routes` includes `\/v1\/skills`, `anthropic_routes`, or `llm_api_routes`, could upload a crafted skill archive containing path traversal entries.\n\nWhen the skill was processed for execution, those entries could be written outside the intended extraction\/staging directory. This could allow arbitrary file write and may lead to code execution depending on deployment configuration and writable paths.\n\n### Patches\n\nThe issue is fixed in `1.83.7-stable`.\n\nLiteLLM recommens upgrading to `1.83.7-stable` or later.\n\n### Workarounds\n\nIf upgrading is not immediately possible:\n\n1. Block `POST \/v1\/skills` at your reverse proxy or API gateway.\n2. Restrict Skills API access to trusted users only.","fixed_in":["1.83.7"],"id":"PYSEC-2026-3477","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-3477","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-59822","GHSA-7488-6r32-c95q"],"details":"### Impact\n\nLiteLLM's MCP Streamable HTTP endpoint could allow an unauthenticated attacker to establish an authenticated MCP session using an arbitrary Bearer token.\n\nThe MCP auth handler supported OAuth2 passthrough for upstream MCP servers, but the fallback path could replace failed LiteLLM key validation with an empty `UserAPIKeyAuth()` object. This allowed requests with a fabricated `Authorization` header to reach MCP tooling without a valid LiteLLM key.\n\nAn attacker could use this to list and call configured MCP tools and access connected services exposed through MCP.\n\n### Patches\n\nThe issue is fixed in `1.84.0`.\n\nWe recommend upgrading to `1.84.0` or later.\n\n### Workarounds\n\nIf upgrading is not immediately possible, disable MCP routes or block access to `\/mcp\/` and related MCP endpoints at your reverse proxy or API gateway.\n\n### References\n\n* [v1.84.0](https:\/\/github.com\/BerriAI\/litellm\/releases\/tag\/v1.84.0)","fixed_in":["1.84.0"],"id":"PYSEC-2026-3479","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-3479","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-59819","GHSA-4g5m-c9r5-49xf"],"details":"### Impact\n\nLiteLLM's `\/health\/test_connection` endpoint resolved request-supplied environment and OIDC file references in `litellm_params`. A proxy administrator, or another privileged caller with permission to test model connections, could cause LiteLLM to read files from the local filesystem via an `oidc\/file\/` reference.\n\nBecause exploitation requires privileged proxy access, this is treated as a defense-in-depth issue rather than a cross-tenant privilege bypass.\n\n### Patches\n\nThe issue is fixed in `1.83.10-stable`.\n\nLiteLLM recommend upgrading to `1.83.10-stable` or later.\n\n### Workarounds\n\nRestrict `\/health\/test_connection` access to trusted administrators only.","fixed_in":["1.83.10"],"id":"PYSEC-2026-3476","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-3476","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-84377"],"details":"### Impact\n\nAny authenticated LiteLLM proxy user could redirect an outbound provider call to a destination they control and cause the proxy to send its own configured provider credentials to that destination. The proxy's request-body validation was a denylist that did not cover every sensitive parameter and did not inspect parameters nested inside other request fields, so a caller could supply a routing or credential value that the proxy applied without clearing the operator's stored key. Any authenticated user could therefore exfiltrate the operator's upstream provider credentials and other configured secrets, and perform Server-Side Request Forgery against internal services reachable from the proxy.\n\n### Patches\n\nFixed in 1.96.2, 1.95.1, 1.94.3, 1.93.2, 1.92.2, 1.91.5, 1.90.7, 1.89.7, and 1.88.6.\n\n### Workarounds\n\nSet `general_settings.allow_client_side_credentials` to `false` so callers cannot override connection parameters, restrict proxy keys to trusted callers, and block the affected parameters (`api_base`, `base_url`, `model_list`, `fallbacks`, provider credential fields) at a reverse proxy or API gateway.","fixed_in":["1.88.6","1.89.7","1.90.7","1.91.5","1.92.2","1.93.2","1.94.3","1.95.1","1.96.2"],"id":"GHSA-3cv6-jpf6-8222","link":"https:\/\/osv.dev\/vulnerability\/GHSA-3cv6-jpf6-8222","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-84377","GHSA-3cv6-jpf6-8222"],"details":"### Impact\n\nAny authenticated LiteLLM proxy user could redirect an outbound provider call to a destination they control and cause the proxy to send its own configured provider credentials to that destination. The proxy's request-body validation was a denylist that did not cover every sensitive parameter and did not inspect parameters nested inside other request fields, so a caller could supply a routing or credential value that the proxy applied without clearing the operator's stored key. Any authenticated user could therefore exfiltrate the operator's upstream provider credentials and other configured secrets, and perform Server-Side Request Forgery against internal services reachable from the proxy.\n\n### Patches\n\nFixed in 1.96.2, 1.95.1, 1.94.3, 1.93.2, 1.92.2, 1.91.5, 1.90.7, 1.89.7, and 1.88.6.\n\n### Workarounds\n\nSet `general_settings.allow_client_side_credentials` to `false` so callers cannot override connection parameters, restrict proxy keys to trusted callers, and block the affected parameters (`api_base`, `base_url`, `model_list`, `fallbacks`, provider credential fields) at a reverse proxy or API gateway.","fixed_in":["1.88.6","1.89.7","1.90.7","1.91.5","1.92.2","1.93.2","1.94.3","1.95.1","1.96.2"],"id":"PYSEC-2026-4066","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-4066","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-37004"],"details":"BerriAI litellm <=1.82.4 is vulnerable to Server-Side Template Injection (SSTI), which allows unauthenticated remote attackers to execute arbitrary OS commands via a crafted dotprompt_content parameter in the \/prompts\/test endpoint due to use of an unsandboxed jinja2.Environment.","fixed_in":["1.83.7"],"id":"GHSA-6wvf-77m9-58rm","link":"https:\/\/osv.dev\/vulnerability\/GHSA-6wvf-77m9-58rm","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-37004","GHSA-6wvf-77m9-58rm"],"details":"BerriAI litellm <=1.82.4 is vulnerable to Server-Side Template Injection (SSTI), which allows unauthenticated remote attackers to execute arbitrary OS commands via a crafted dotprompt_content parameter in the \/prompts\/test endpoint due to use of an unsandboxed jinja2.Environment.","fixed_in":["1.83.7"],"id":"PYSEC-2026-3861","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-3861","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-12773","GHSA-4jcj-7x88-m979"],"details":"A weakness has been identified in BerriAI litellm up to 1.59.8. Affected is the function UserAPIKeyAuth of the file litellm\/proxy\/_experimental\/mcp_server\/auth\/user_api_key_auth_mcp.py of the component MCP Proxy. Executing a manipulation can lead to improper authentication. The attack may be launched remotely. The exploit has been made available to the public and could be used for attacks. The vendor was contacted early about this disclosure.","fixed_in":["1.84.0"],"id":"PYSEC-2026-4067","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-4067","source":"osv","summary":null,"withdrawn":null},{"aliases":["CVE-2026-59823","GHSA-hx8v-g79f-8w5f"],"details":"### Summary\nA server-side request forgery in LiteLLM Proxy lets an authenticated caller redirect the\nproxy's outbound request to a host of their choosing by smuggling an `api_base` inside the\n`user_config` request body, bypassing the existing parameter guard.\n\n### Details\nLiteLLM Proxy validates request bodies with `is_request_body_safe`, which blocks the\n`api_base` and `base_url` parameters but does not cover `user_config`. The `user_config`\nobject is used to build the outbound router for a request, so a caller can place an\n`api_base` inside it and reach an arbitrary host. The guard only inspected the two\ntop-level keys, so the same `api_base` nested inside `user_config` was never checked.\n\nExploitation requires a valid virtual key.\n\n### Impact\nAn authenticated caller can make the proxy issue server-side requests to internal or\nexternal hosts of their choosing, reaching endpoints the caller cannot otherwise access.\n\n### Affected \/ Patched\nAffected: `<= 1.83.8`\nPatched: `1.83.9`\n\n### Remediation\nUpgrade to 1.83.9 or later (released 2026-04-17).","fixed_in":["1.83.9"],"id":"PYSEC-2026-4070","link":"https:\/\/osv.dev\/vulnerability\/PYSEC-2026-4070","source":"osv","summary":null,"withdrawn":null},{"aliases":[],"details":"### Summary\n\nThe `\/sso\/debug\/callback` endpoint reflected identity-provider-controlled fields into an inline `<script>` block via `json.dumps`, which escapes JSON string syntax but not HTML context characters such as `<\/script>`. An IdP-returned string field containing `<\/script>` (for example a user-editable display name) closed the script block and executed attacker JavaScript in the proxy's origin.\n\n### Impact\n\nReflected XSS on the LiteLLM proxy origin. The route required no authentication, so an attacker with an account on the configured IdP whose claims pass such strings through verbatim could complete the SSO flow themselves and deliver the resulting callback URL to a logged-in proxy admin. Scripts executing in the proxy origin can read the dashboard session cookie and act with the victim's privileges, up to full admin account takeover.","fixed_in":["1.85.0"],"id":"GHSA-hhww-mrg2-969h","link":"https:\/\/osv.dev\/vulnerability\/GHSA-hhww-mrg2-969h","source":"osv","summary":null,"withdrawn":null},{"aliases":[],"details":"## Summary\n\nAn authenticated LiteLLM proxy user could read arbitrary local files from the proxy runtime by sending a crafted `vertex_ai_credentials` value to `\/utils\/transform_request`. The request-body safety blocklist rejected `vertex_credentials` but missed its alias `vertex_ai_credentials`, which let the value reach the Vertex AI credential loader and trigger a file read that is sent to an attacker-controlled URL.\n\n## Impact\n\nA low-privilege authenticated user (only `internal_user_viewer`, obtainable through SSO login on many deployments) could read any file readable by the LiteLLM process. Reading files such as `\/proc\/self\/environ` can disclose `LITELLM_MASTER_KEY` and provider API keys, so the impact can escalate to full proxy-admin control.\n\n## Remediation\n\nUpgrade to `1.95.0` or later. The fix adds the missing alias to the request-body blocklist, so no configuration change is needed once you upgrade. There is no reliable configuration workaround on affected versions.","fixed_in":["1.95.0"],"id":"GHSA-g5ff-637f-6q2m","link":"https:\/\/osv.dev\/vulnerability\/GHSA-g5ff-637f-6q2m","source":"osv","summary":null,"withdrawn":null}]}