Skip to main content
Arka is secure by default. Symbolic checks in arka_security.py run before web search, risky actions, and destructive shell commands.
Security is on by default. Set SECURITY=0 only if you understand the risks.

Security layers

Disable individual layers by setting them to 0:

What gets blocked

Prompt injection — requests that try to override system instructions before web search:
Destructive shell — Arka hard-blocks patterns that could damage your system, with no override:
Risky actions — installs, deletes, downloads, and automation require confirmation:

Local execution model

Commands run locally on your machine. LLM calls go through the fallback orchestrator across 24 providers — the model never executes shell commands directly in the cloud.
No Docker required for daily use. PrivateGPT and Qdrant are only needed for PDF ingest.

Goal agent safety

The autonomous goal agent (arka goal) has a step budget (GOAL_MAX_STEPS=25 by default) and respects the same security gates. Use -y to auto-confirm risky steps only when you trust the environment.