Skip to main content
drako init creates a .drako.yaml configuration file in the current directory. It reads the cached scan results from a prior drako scan run (or runs the scanner automatically) and uses the discovered agents, tools, and models to produce a config tailored to your project.

Usage

drako init [OPTIONS]

Options

--api-key
string
Drako API key. If omitted, the command reads DRAKO_API_KEY from the environment, or prompts you to enter one. Get a free key at app.getdrako.com/signup.
--framework
crewai | langgraph | autogen | generic
Override framework auto-detection. When omitted, Drako detects the framework from your project files.
--endpoint
string
default:"https://api.getdrako.com"
Custom backend endpoint.
--autopilot
flag
Generate config with smart defaults derived from your scan results. All features start in audit mode — logging, not blocking. This is the default when no mode flag is passed.
--balanced
flag
Generate config with enforcement active and escape hatches enabled. DLP and ODD switch to enforce mode; HITL times out with reject.
--strict
flag
Generate config with maximum governance for enterprise use. Every feature is enforced, intent verification is enabled with anti-replay, and the audit trail is cryptographic with 90-day retention.
--manual
flag
Generate a full YAML template with every section and commented-out options. Intended for teams who want complete control and prefer to edit the config by hand.
--template
string
Start from an industry policy template (e.g. fintech, healthcare). The template is merged with the generated config; your project-specific values take precedence. See drako templates list for available names.

Governance levels

Smart defaults derived from your scan. Everything starts in audit mode — observe before enforcing.
  • ODD: each agent locked to its discovered tools
  • DLP: audit mode (logging PII, not blocking)
  • Circuit breaker: threshold 5 failures / 60 s window
  • HITL: active for write/execute tools, auto-allow on timeout
  • Audit: enabled
  • FinOps: cost tracking enabled
drako init
# or explicitly:
drako init --autopilot

What gets generated

The generated .drako.yaml contains:
  • Discovered agents with their source file paths
  • Discovered tools with inferred types (read, network, write, execute)
  • ODD (Operational Design Domain) boundaries per agent
  • Magnitude limits computed from agent/tool count and model cost tier
  • DLP, circuit breaker, HITL, audit, and FinOps sections
If .drako.yaml already exists, you are prompted to overwrite, merge new tools, or skip.

Examples

# Run scan first to populate cache, then generate config
drako scan .
drako init

# Skip scan prompt if you have a fresh cache
drako init --autopilot

# Start with enforcement active
drako init --balanced

# Enterprise config
drako init --strict

# Full template for manual editing
drako init --manual

# Industry template
drako init --template fintech

# Combine governance level with a template
drako init --balanced --template healthcare

# Override framework detection
drako init --framework langgraph
After drako init, run drako push to sync the config to the Drako platform, then add govern() to your agent code to activate runtime enforcement.

Build docs developers (and LLMs) love