Docs

Quickstart

Five minutes, three steps: install, construct an agent, stream a turn. This walkthrough uses the Node adapter — the browser version is the same shape with a different import.

1. Install#

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

2. Construct an agent#

An agent is a model plus a place to work. The Node adapter points it at a real directory and pre-wires the file tools and a sandboxed shell.

agent.ts

import { NodeAgent } from '@nanocodana/nodejs'
import { anthropic } from '@ai-sdk/anthropic'

const agent = NodeAgent({
  model: anthropic('claude-sonnet-5'),
  workingDirectory: process.cwd(),
  // Risky tools prompt for approval; everything else runs freely.
  needsApproval: ['Write', 'Edit', 'MultiEdit', 'Delete'],
})

3. Stream a turn#

agent.ts

const messages = [{ role: 'user', content: 'Rename getFoo to getBar across src/.' }]
const result = await agent.stream({ messages })

for await (const chunk of result.fullStream) {
  if (chunk.type === 'text-delta') process.stdout.write(chunk.text)
  else if (chunk.type === 'tool-call') console.log(`\n↳ ${chunk.toolName}`, chunk.input)
  else if (chunk.type === 'tool-approval-request') {
    // Decide, then continue the turn — see the Tool Approval docs.
  }
}

Run it with npx tsx agent.ts (or compile with tsc) and watch the agent read your project, propose edits, and ask before touching anything you gated.

Where to next#

Streaming & Results — every chunk type and how to keep conversation history.
Tool Approval — wire the human-in-the-loop decision into your UI.
Agent Skills — teach the agent your team's conventions with SKILL.md files.

Prefer zero code? npm install -g @nanocodana/cli and run codana chat in any project — it's this exact agent in a terminal UI.