Configuration

All configuration is stored in ~/.guardclaw/.env. You can edit it via the CLI or the dashboard Settings panel — no manual file editing needed.

CLI Configuration

guardclaw config llm          # interactive LLM backend picker
guardclaw config mode         # set approval mode (auto / prompt / monitor-only)
guardclaw config thresholds   # set risk score thresholds
guardclaw config show         # print current config
guardclaw config set KEY VALUE  # set any env var, applied immediately

All config commands apply changes to the running server instantly — no restart needed.

Approval Modes

Mode	Behavior
auto	Auto-allow score ≤ 6, auto-block score ≥ 9, warn on 7–8
prompt	Ask user for approval on score 7–8
monitor-only	Never block, just log everything

Set via CLI:

guardclaw config mode
# or directly:
guardclaw config set GUARDCLAW_APPROVAL_MODE prompt

Risk Thresholds

Variable	Default	Meaning
GUARDCLAW_AUTO_ALLOW_THRESHOLD	6	Score ≤ this → auto-allow
GUARDCLAW_ASK_THRESHOLD	8	Score ≤ this → ask user (prompt mode)
GUARDCLAW_AUTO_BLOCK_THRESHOLD	9	Score ≥ this → auto-block

guardclaw config thresholds

Environment Variables

Full list of supported variables:

# LLM Backend
SAFEGUARD_BACKEND=lmstudio    # lmstudio | ollama | anthropic | built-in | openrouter | fallback

# LM Studio
LMSTUDIO_URL=http://localhost:1234/v1
LMSTUDIO_MODEL=auto
LMSTUDIO_API_KEY=             # optional, for hosted OpenAI-compatible endpoints

# Ollama
OLLAMA_URL=http://localhost:11434
OLLAMA_MODEL=llama3

# Anthropic
ANTHROPIC_API_KEY=sk-ant-...

# OpenRouter
OPENROUTER_API_KEY=sk-or-...
OPENROUTER_MODEL=openai/gpt-4o

# Approval
GUARDCLAW_APPROVAL_MODE=auto
GUARDCLAW_AUTO_ALLOW_THRESHOLD=6
GUARDCLAW_ASK_THRESHOLD=8
GUARDCLAW_AUTO_BLOCK_THRESHOLD=9

# Server
PORT=3002

Dashboard Settings Panel

Open the ⚙️ Settings panel in the dashboard to:

• Switch LLM backend and model
• Test connection to the backend
• Toggle fail-closed mode
• Manage whitelist/blacklist

Changes apply instantly without restart.