Start Here

Updated 17 minutes ago • August 2, 2026

Kortyx is an opinionated framework for teams building production AI features that need explicit behavior, predictable runtime flow, and strong control over providers, streaming, and state.

Choose your path

• New project setup: Installation
• First end-to-end flow (recommended for live chunking): Quickstart (Next.js API Route)
• Server Action variant (buffered return, no live chunk UI): Quickstart (Next.js Server Action)
• Build mental model first: Define Workflows, Nodes, and Runtime Context
• Implement task-oriented features: Project Structure, Rendering Streamed Chat, and Interrupts and Resume
• Inspect your application locally: Run Kortyx Studio Locally
• Connect or deploy Studio: Studio Overview
• Harden for production: Production
• Verify exact APIs and protocols: Reference

Stack direction

• Next.js: fastest path for an initial app integration.
• Express/Fastify: strong fit for server-heavy orchestration and infrastructure-owned deployments.
  • Start with React Frontend + Node Backend
  • Then read streamChat
  • Then configure Runtime Persistence Adapters

Good to know: The docs are TypeScript-first and prioritize system-builder workflows with production reliability in mind.

Observe your Kortyx application

Kortyx Studio turns server-side telemetry into readable sessions, runs, workflow transitions, model calls, interrupts, timing, token usage, and cost.

If you already have a Kortyx agent running, start the complete local stack with one command:

Then follow Run Kortyx Studio Locally to connect the agent and inspect its first run. Studio is optional observability infrastructure; it does not execute your workflows, and your application continues running if Studio is unavailable.