Environment Variables

cascadeflow reads configuration from environment variables as part of its resolution chain: Code > Environment Variables > Config Files > Defaults.

Harness Configuration

Variable	Type	Description
`CASCADEFLOW_HARNESS_MODE`	`off \| observe \| enforce`	Harness activation mode
`CASCADEFLOW_HARNESS_BUDGET`	`float`	Budget limit in USD
`CASCADEFLOW_HARNESS_MAX_TOOL_CALLS`	`int`	Maximum tool calls allowed
`CASCADEFLOW_HARNESS_MAX_LATENCY_MS`	`float`	Maximum latency in milliseconds
`CASCADEFLOW_HARNESS_COMPLIANCE`	`gdpr \| hipaa \| pci \| strict`	Compliance mode
`CASCADEFLOW_CONFIG`	`path`	Path to config file

Provider API Keys

Variable	Provider
`OPENAI_API_KEY`	OpenAI
`ANTHROPIC_API_KEY`	Anthropic
`GROQ_API_KEY`	Groq
`TOGETHER_API_KEY`	Together AI
`OPENROUTER_API_KEY`	OpenRouter
`HUGGINGFACE_API_KEY`	Hugging Face

Resolution Order

When cascadeflow.init() is called, settings resolve in this order (first wins):

Code — arguments passed directly to init(), run(), or @agent()
Environment — CASCADEFLOW_HARNESS_* variables
Config file — path from CASCADEFLOW_CONFIG or default locations
Defaults — mode=off, no budget, no compliance

Example

# .env
CASCADEFLOW_HARNESS_MODE=observe
CASCADEFLOW_HARNESS_BUDGET=1.00
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...

import cascadeflow

# Picks up mode="observe" and budget=1.00 from env
report = cascadeflow.init()
print(report.config_sources)  # Shows where each setting came from

Python

TypeScript

Environment Variables

Environment Variables

Harness Configuration

Provider API Keys

Resolution Order

Example

​Environment Variables

​Harness Configuration

​Provider API Keys

​Resolution Order

​Example

Environment Variables

Harness Configuration

Provider API Keys

Resolution Order

Example