NovalinkIn dev

Blog

Introducing Novalink: AI agent workflows you can watch run

By Oshadha Vimukthi ·

Most agent frameworks hide what happened inside a run. Novalink puts every step on a canvas: you see which node is running, what it received, what it returned, and which branch was skipped, then and later.

Why another workflow tool

When an agent workflow gives a wrong answer, the question is always the same: which step went wrong, and with what input? Logs answer that slowly. A canvas that replays the run answers it at a glance.

Novalink keeps the system small on purpose. There is no catalogue of hundreds of integrations: a handful of nodes, plus one HTTP Request node that works as a step or as a tool an agent can call.

What it does today

  • A canvas for real graphs: triggers, Claude agents with tools and structured output, branching with If and Switch, and Merge.
  • Live runs: every node streams its status to the canvas as the run executes.
  • Replay: each run stores the exact graph it executed, so you can reopen it after the workflow has changed.
  • Versions: publishing freezes a draft, so editing never changes what a published version runs.
  • Generate with AI: describe a workflow and Claude drafts a validated graph you preview before creating.
  • Hosted or self-hosted: use it at novalink.live, or run it on your own machine.

What is next

Novalink is under active development. The next phase brings the webhook endpoint, scheduled triggers, and a worker queue so runs survive restarts. Follow along on GitHub.

Try it

Start with the getting started guide, or read one of the use cases and build it yourself. Questions and feedback go to support@novalink.live.

Build it on the canvas

Create a free account and run your first workflow in the browser, or run Novalink on your own machine.