BoringLLM

A tiny, stateless API that gives OpenAI, Claude, Gemini, xAI and Ollama one simple text-in/text-out contract.

Send a text message to a model, get text back. The provider is picked from the model name. That is the whole product.

POST /v1/chat

What it deliberately does not do

No streaming. No conversation history. No system prompts. No tool calling. No structured output. No images, audio or embeddings. No temperature, max tokens or other generation options. No retries, fallbacks or routing. No caching, rate limiting, usage tracking or dashboard. No database. No authentication.

BoringLLM is a compatibility layer, not an AI gateway. If you need those things, put them in front of it or use something bigger.

It also stores nothing: requests and responses are never written to disk or memory beyond the life of the request. What each provider does with the text it receives is governed by that provider's own data handling and retention policy.

Supported providers

Prefix	Provider	Credential
openai	OpenAI	OPENAI_API_KEY
anthropic	Anthropic / Claude	ANTHROPIC_API_KEY
gemini	Google Gemini	GEMINI_API_KEY
xai	xAI	XAI_API_KEY
ollama	Ollama	OLLAMA_BASE_URL

Only the providers you actually call need to be configured.

Model naming

<provider>/<model>

The part before the first / selects the provider; everything after it is sent to that provider unchanged, so model names containing slashes or tags still work.

openai/gpt-5
anthropic/claude-sonnet-4-5
gemini/gemini-2.5-pro
xai/grok-4
ollama/qwen3
ollama/hf.co/user/model:q4_k_m

Quick start

Requires Node.js 24 or newer.

npm install
cp .env.sample .env    # add at least one key
npm run dev            # or: npm run build && npm start

Environment variables

Variable	Default	Notes
PORT	3000	HTTP port
OPENAI_API_KEY	—	Required for openai/*
ANTHROPIC_API_KEY	—	Required for anthropic/*
GEMINI_API_KEY	—	Required for gemini/*
XAI_API_KEY	—	Required for xai/*
OLLAMA_BASE_URL	http://localhost:11434	Required for ollama/*

Docker

cp .env.sample .env
docker compose up --build

Compose reads your .env. Ollama is not bundled: set OLLAMA_BASE_URL to an Ollama instance you already run. From inside the container use http://host.docker.internal:11434 rather than localhost.

Request

curl -s localhost:3000/v1/chat \
  -H 'Content-Type: application/json' \
  -d '{"model":"anthropic/claude-sonnet-4-5","input":"Explain dependency injection in one paragraph."}'

{
  "model": "anthropic/claude-sonnet-4-5",
  "input": "Explain dependency injection in one paragraph."
}

Response

{
  "model": "anthropic/claude-sonnet-4-5",
  "output": "Dependency injection is ..."
}

Errors use one small shape, whichever provider failed:

{
  "error": {
    "code": "UNSUPPORTED_PROVIDER",
    "message": "Unsupported model provider: example"
  }
}

Code	Status
INVALID_REQUEST	400 (413 if the body exceeds 1 MB)
UNSUPPORTED_PROVIDER	400
NOT_FOUND	404
PROVIDER_NOT_CONFIGURED	500
INTERNAL_ERROR	500
PROVIDER_ERROR	502

GET /health returns {"status":"ok"}.

Security

There is no authentication, by design. Run BoringLLM as an internal service, bound to a private network, and add authentication in front of it if you expose it publicly.

License

MIT