Skip to content

Getting Started

Varun S V edited this page Oct 16, 2026 · 1 revision

Getting Started

Prerequisites

  • Node.js 22.22.2 or later (required by the root package.json).
  • An accessible model provider: Anthropic, OpenAI, Gemini, Azure, Ollama, or another OpenAI-compatible endpoint.
  • A terminal and a project folder that you control. For self-hosting, install Docker with Compose instead; see Cloud-and-Self-Hosting.

Install and initialize

npm install -g cascade-ai
mkdir my-project
cd my-project
cascade init
cascade doctor
cascade

Inside Cascade's interactive REPL, type /model to select provider → tier → model, or use Auto routing. Try an exploratory task before allowing writes:

Explain the organization of this codebase. Identify the most important modules without changing anything.

For a noninteractive run:

cascade run "Explain the authentication design of the current workspace"

Provider configuration

Set credentials using a supported environment variable, or configure providers in .cascade/config.json. Avoid committing provider keys.

Provider Common setup
Anthropic ANTHROPIC_API_KEY
OpenAI OPENAI_API_KEY
Gemini GOOGLE_API_KEY (credential linking also recognizes some Gemini aliases)
Azure OpenAI Azure resource endpoint, deployment name, API version/key
Ollama Local server, normally http://localhost:11434
OpenAI-compatible Explicit baseUrl, model, key if needed, and correct local cost designation

For manual config, see src/config/schema.ts. cascade doctor diagnoses common problems. /why explains routing, and /cost shows token/cost tracking. See Models-and-Budgets.

Choose another interface

First-run safety

  1. Point file and shell tools only at a workspace you intend to modify.
  2. Require explicit approval for dangerous operations while evaluating the platform.
  3. Review the model and cost settings before a large task.
  4. Never expose CLOUD_DEV_BYPASS=1 beyond a loopback-only local test server.

Sources: README, CLI entry.

Clone this wiki locally