Configuration

Crew reads settings from three places. The first one that has a value wins:

1. Environment variables
2. Settings saved in the database (web UI Settings, or crew config set)
3. Defaults

Env	Setting	Default
CREW_HOME	none	~/.crew (the database is crew.db)
CREW_PORT	none	3000
CREW_HOST	none	127.0.0.1
CREW_PROVIDER	provider	ollama
OLLAMA_HOST	ollamaHost	http://127.0.0.1:11434
OPENAI_BASE_URL	openaiBaseUrl	https://api.openai.com/v1
OPENAI_API_KEY	openaiApiKey	empty
CREW_MODEL	model	qwen2.5:3b
CREW_EMBED_MODEL	embedModel	nomic-embed-text
CREW_TELEGRAM_TOKEN	telegramToken	empty

Providers

Ollama is the default. If it runs locally, there is nothing to configure.

Anything that speaks the OpenAI API also works: LM Studio, llama.cpp server, vLLM, OpenRouter, OpenAI.

CREW_PROVIDER=openai OPENAI_BASE_URL=http://127.0.0.1:1234/v1 crew start

Security notes

Crew binds to 127.0.0.1 by default. The API has no authentication, so don't expose it to the internet without a reverse proxy that adds it.

Secrets are masked in the settings API.

All state is in ~/.crew/crew.db. To back it up, copy the file.