Documentation menu

Docs

Installation

Pick the package for where your agent runs, then add the AI SDK provider for whichever model you use. Node 18+ for the Node packages; any modern browser for the browser package.

Node#

npm install @nanocodana/nodejs @ai-sdk/anthropic ai

Browser#

npm install @nanocodana/browser @ai-sdk/openai ai

The terminal app#

npm install -g @nanocodana/cli   # provides the `codana` command

Providers#

The model is always yours to choose — install the AI SDK provider package for it alongside NanoCodana: @ai-sdk/anthropic, @ai-sdk/openai, @ai-sdk/google, or @ai-sdk/openai-compatible for OpenRouter, DeepSeek, Together, Groq, Ollama, LM Studio, and friends.

Size#

What you deploy is small: a bundled Node agent is 3.58 MB, a browser app loads 182 kB gzipped up front, and the CLI is a single 6.2 MB file. What you see in node_modules is bigger — 141 MB on Node — because the shell carries wasm runtimes it loads on demand. That's a build-time cost; bundling ships only what you import, and each package documents how to prune the rest. See Node, core, or the CLI.

Embedding with a custom filesystem or sandbox? Install @nanocodana/core directly — the adapters are thin wrappers over it.