Environment Variables Reference

All Environment Variables

Variable	Default	Description
`OPENCLAW_PROVIDER`	openai	Default LLM provider (openai, claude, ollama, gemini)
`OPENCLAW_MODEL`	gpt-4o	Default model name
`OPENCLAW_OPENAI_KEY`	(none)	OpenAI API key
`OPENCLAW_CLAUDE_KEY`	(none)	Anthropic Claude API key
`OPENCLAW_GEMINI_KEY`	(none)	Google Gemini API key
`OPENCLAW_GROK_KEY`	(none)	xAI Grok API key
`OPENCLAW_OLLAMA_URL`	http://localhost:11434	Ollama server base URL
`OPENCLAW_MAX_STEPS`	20	Default maximum agent execution steps
`OPENCLAW_TIMEOUT`	300	Default task timeout in seconds
`OPENCLAW_LOG_LEVEL`	info	Logging verbosity (debug, info, warn, error)
`OPENCLAW_CONFIG_PATH`	~/.openclaw/config.yaml	Path to config file
`OPENCLAW_PROFILE`	default	Active configuration profile name
`OPENCLAW_NO_COLOR`	false	Disable ANSI color in terminal output
`OPENCLAW_ALLOWED_TOOLS`	all	Comma-separated list of allowed tool packs

Priority Order

When the same setting is defined in multiple places, OpenClaw uses this priority (highest first):

CLI flags — e.g., --provider claude
Environment variables — e.g., OPENCLAW_PROVIDER=claude
Config file values — e.g., ~/.openclaw/config.yaml
Built-in defaults — hardcoded fallback values

💡

Use This for Environment Overrides

This layered approach means you can run OPENCLAW_PROVIDER=ollama openclaw run "..." to use Ollama for just one command without changing your config file.

.env File Support

OpenClaw automatically loads a .env file from the current working directory if one exists:

# .env file in your project
OPENCLAW_PROVIDER=openai
OPENCLAW_OPENAI_KEY=sk-proj-abc123...
OPENCLAW_MODEL=gpt-4o
OPENCLAW_MAX_STEPS=25
OPENCLAW_LOG_LEVEL=debug

# .gitignore — always exclude .env files
echo ".env" >> .gitignore

Docker & Container Setup

Pass environment variables when running OpenClaw in a container:

# docker run with env vars
docker run --rm \
  -e OPENCLAW_PROVIDER=openai \
  -e OPENCLAW_OPENAI_KEY="sk-proj-..." \
  -e OPENCLAW_MODEL=gpt-4o \
  -v $(pwd):/workspace \
  openclaw/openclaw:latest \
  run "Analyze the codebase" --context /workspace

# docker-compose.yml
services:
  openclaw:
    image: openclaw/openclaw:latest
    environment:
      - OPENCLAW_PROVIDER=openai
      - OPENCLAW_OPENAI_KEY=${OPENAI_API_KEY}
      - OPENCLAW_MODEL=gpt-4o
    volumes:
      - ./workspace:/workspace

💡

See Also

API Keys Setup — provider-specific key configuration
openclaw config command
Config Examples — ready-to-use configurations

Setting Variables Per Operating System

Linux / macOS

# Add to ~/.bashrc or ~/.zshrc for persistence
export OPENCLAW_OPENAI_KEY="sk-proj-..."
export OPENCLAW_DEFAULT_MODEL="gpt-4o-mini"
export OPENCLAW_LOG_LEVEL="info"

# Apply immediately
source ~/.bashrc

# Temporary (current session only)
OPENCLAW_LOG_LEVEL=debug openclaw run "my task"

Windows (PowerShell)

# Current session
$env:OPENCLAW_OPENAI_KEY = "sk-proj-..."
$env:OPENCLAW_DEFAULT_MODEL = "gpt-4o-mini"

# Persistent (user-level, survives reboots)
[System.Environment]::SetEnvironmentVariable(
    "OPENCLAW_OPENAI_KEY", "sk-proj-...",
    [System.EnvironmentVariableTarget]::User
)

# Single-run override
$env:OPENCLAW_LOG_LEVEL="debug"; openclaw run "my task"

Docker

# Dockerfile
ENV OPENCLAW_DEFAULT_PROVIDER=openai
ENV OPENCLAW_DEFAULT_MODEL=gpt-4o-mini
ENV OPENCLAW_LOG_LEVEL=info
# Do NOT bake secret keys into Dockerfiles — use --env-file or secrets

# docker run with env file (keep .env out of version control)
docker run --env-file .env my-openclaw-image openclaw run "task"

Full Environment Variable Reference

Variable	Default	Description
`OPENCLAW_OPENAI_KEY`	—	OpenAI API key
`OPENCLAW_ANTHROPIC_KEY`	—	Anthropic (Claude) API key
`OPENCLAW_OLLAMA_URL`	`http://localhost:11434`	Ollama base URL
`OPENCLAW_DEFAULT_PROVIDER`	`openai`	Default LLM provider
`OPENCLAW_DEFAULT_MODEL`	`gpt-4o-mini`	Default model name
`OPENCLAW_MAX_STEPS`	`10`	Max autonomous steps per task
`OPENCLAW_TIMEOUT`	`120`	API call timeout in seconds
`OPENCLAW_LOG_LEVEL`	`warning`	`debug`, `info`, `warning`, `error`
`OPENCLAW_CONFIG_PATH`	`~/.openclaw/config.yaml`	Override config file location
`OPENCLAW_DATA_DIR`	`~/.local/share/openclaw`	Logs and state storage directory
`OPENCLAW_ALLOW_EXEC`	`true`	Allow shell command execution by agent
`HTTPS_PROXY`	—	HTTP/S proxy for all API calls

Priority Order

OpenClaw resolves configuration in this order (higher wins):

CLI flags (--model gpt-4o)
Environment variables (OPENCLAW_DEFAULT_MODEL)
Project-level .openclaw.yaml in current directory
User config ~/.openclaw/config.yaml
Built-in defaults

.env File Support

OpenClaw automatically loads a .env file in the current directory if it exists. This is the most convenient method for project-specific secrets:

# myproject/.env
OPENCLAW_OPENAI_KEY=sk-proj-...
OPENCLAW_DEFAULT_MODEL=gpt-4o
OPENCLAW_LOG_LEVEL=info
OPENCLAW_MAX_STEPS=8

# .gitignore — NEVER commit .env files
echo ".env" >> .gitignore
echo ".env.*" >> .gitignore
echo "!.env.example" >> .gitignore   # commit the example without real values

.env.example Template

Check in a sanitized example so team members know which variables to set:

# .env.example  (committed to git — no real values)
OPENCLAW_OPENAI_KEY=your-openai-api-key-here
OPENCLAW_ANTHROPIC_KEY=your-anthropic-key-here
OPENCLAW_DEFAULT_PROVIDER=openai
OPENCLAW_DEFAULT_MODEL=gpt-4o-mini
OPENCLAW_MAX_STEPS=10
OPENCLAW_TIMEOUT=120

# New team member setup:
cp .env.example .env
nano .env  # fill in real values

Per-Run Variable Override

Override any variable for a single command without changing your config or .env file:

# Linux / macOS: prefix variable before the command
OPENCLAW_DEFAULT_MODEL=claude-3-5-sonnet openclaw run "Review this code"
OPENCLAW_LOG_LEVEL=debug openclaw run "Debug this task"

# Windows PowerShell: set then run
$env:OPENCLAW_DEFAULT_MODEL = "claude-3-5-sonnet"
openclaw run "Review this code"
# Reset after:
Remove-Item Env:OPENCLAW_DEFAULT_MODEL

← Local LLM Models Config Examples →