openclaw config — Manage Configuration

Synopsis

openclaw config <subcommand> [KEY] [VALUE]

The config command is your primary interface for managing OpenClaw settings stored in ~/.openclaw/config.yaml. Changes take effect immediately for all subsequent openclaw run and openclaw start calls.

Subcommands

Subcommand	Usage	Description
`get`	`config get KEY`	Print the value of a specific config key
`set`	`config set KEY VALUE`	Set a config key to a value
`list`	`config list`	Print all current configuration keys and values
`reset`	`config reset [--confirm]`	Reset all settings to factory defaults
`export`	`config export [FILE]`	Export current config to a YAML file
`import`	`config import FILE`	Import configuration from a YAML file

Examples

Set Configuration Values

# Set OpenAI API key (stored securely in config)
openclaw config set openai.api_key "sk-proj-..."

# Switch provider to Anthropic Claude
openclaw config set provider claude
openclaw config set claude.api_key "sk-ant-..."

# Change default model
openclaw config set model claude-3-5-sonnet

# Set max steps globally
openclaw config set max_steps 30

# Set log level
openclaw config set log_level debug

Read Configuration Values

# Get the current provider
openclaw config get provider
# Output: openai

# Check which model is configured
openclaw config get model
# Output: gpt-4o

# Get Ollama base URL (if configured)
openclaw config get ollama.base_url
# Output: http://localhost:11434

List All Settings

openclaw config list

provider: openai
model: gpt-4o
max_steps: 20
timeout: 300
log_level: info
openai:
  api_key: sk-proj-***redacted***
  model: gpt-4o
claude:
  api_key: (not set)
ollama:
  base_url: http://localhost:11434
  model: llama3.2

Reset to Defaults

openclaw config reset --confirm
# All settings reset to defaults. API keys removed.

Export and Import

# Export current config
openclaw config export ~/openclaw-backup.yaml

# Import on a new machine
openclaw config import ~/openclaw-backup.yaml

Config File Location

Configuration is stored in ~/.openclaw/config.yaml on all platforms:

OS	Default Path
Linux / macOS	`~/.openclaw/config.yaml`
Windows	`%USERPROFILE%\.openclaw\config.yaml`

You can override this location with OPENCLAW_CONFIG_PATH:

export OPENCLAW_CONFIG_PATH=/etc/openclaw/config.yaml

Environment Variable Overrides

Any config key can be overridden by an environment variable at runtime without modifying the config file. Environment variables take precedence over values in config.yaml:

# Override provider for a single session
OPENCLAW_PROVIDER=ollama openclaw run "Analyze this file" --context app.py

# PowerShell equivalent
$env:OPENCLAW_PROVIDER = "ollama"
openclaw run "Analyze this file" --context app.py

💡

Priority Order

CLI flags > Environment variables > Config file values > Built-in defaults. This allows layered configuration for different environments.

💡

See Also

API Keys Setup — provider-specific key configuration
Configuration Overview
Environment Variables Reference

Config Profiles

Maintain separate profiles for different projects or environments. Switch instantly with a single command or environment variable:

# Create profiles by exporting different configs to named files
openclaw config set provider openai
openclaw config set openai.model gpt-4o
openclaw config export ~/.openclaw/profiles/work.yaml

openclaw config set provider ollama
openclaw config set model llama3.2
openclaw config export ~/.openclaw/profiles/personal.yaml

# Switch to work profile
openclaw config import ~/.openclaw/profiles/work.yaml

# Or override for a single command:
OPENCLAW_CONFIG_PATH=~/.openclaw/profiles/personal.yaml openclaw run "Draft a short email"

Per-Project Config Override

Place a .openclaw.yaml file in a project directory to use project-specific settings automatically:

# .openclaw.yaml (project root)
provider: openai
openai:
  model: gpt-4o
  max_tokens: 2048
max_steps: 10
allowed_paths:
  - ./src/
  - ./tests/

When you run openclaw from inside that directory, the local .openclaw.yaml overrides ~/.openclaw/config.yaml. Useful for enforcing per-project model choices or path restrictions.

Full Config Reference

# ~/.openclaw/config.yaml — all options with defaults shown
provider: openai          # openai | claude | ollama | gemini
model: gpt-4o             # default model for the provider
max_steps: 20             # max autonomous steps before pausing
timeout: 300              # request timeout in seconds
temperature: 0.7          # LLM temperature (0.0-2.0)
max_tokens: 4096          # max tokens per LLM response
log_level: info           # debug | info | warn | error
log_file: ""              # path to log file; empty = stderr only
dry_run: false            # if true, plan but don't execute
allowed_tools: []         # empty = all tools allowed

openai:
  api_key: ""             # or set OPENAI_API_KEY env var
  base_url: ""            # custom endpoint (Azure, proxy)

claude:
  api_key: ""             # or set ANTHROPIC_API_KEY env var

ollama:
  base_url: "http://localhost:11434"
  model: llama3.2

← openclaw stop openclaw install →