What is neonloops?

What does neonloops do?

Most real work is a mix of two things: steps that must happen the same way every time, and steps nobody can fully script — reading a messy inbox, researching a company, drafting a reply, reconciling two spreadsheets. neonloops lets you draw the first kind as a visual workflow and hand the second kind to AI agents that work inside it. The workflow decides when work starts, what happens in what order, and where a person has to approve. The agents handle the part that changes every time.

You don't have to be in the browser to use it, either — neonloops has an MCP server, so an AI client you already use can build, run and check on workflows for you.

Who is it for?

Operators and small teams whose work is unpredictable, multi-step and nobody's favourite job — operations, support, finance admin, research, reporting. You do not need to write code. You describe the job, neonloops drafts the workflow, and you adjust it on a canvas.

How does it work?

1. Describe the job in plain language, or start from a blank canvas.
2. Shape the workflow: a trigger (a schedule, a webhook, an event in a connected app), the steps, and approval points where a person signs off.
3. Agents do the open-ended steps — with the tools, files and connected apps you gave them — for as long as the job takes.
4. Review everything afterwards: each run keeps a full record of what every step and every agent did.

Which AI models does it use?

neonloops is multi-model for the model steps: you can mix OpenAI, Anthropic and Google inside the same workflow, and there is no API key to supply — we run the provider accounts for you.

Agent steps are the exception, on purpose: the long-running agent work runs on Claude's managed agent infrastructure, so agent steps use Claude models — that's the trade for an agent that keeps working for hours instead of answering once.

How is it priced?

neonloops bills for the time your work is actually working — the minutes a workflow or an agent spends genuinely doing something for you. Not per seat, not per run, and there are no tiers to pick between.

Waiting doesn't bill, and there's no trial: a workspace adds a card and is priced per billable minute, against a ceiling you set yourself.

Does neonloops have an API or SDK?

Not yet. neonloops today is a web app. There is no public API and no SDK for the current product.

What are the neonloops packages on PyPI and npm?

They belong to neonloops v1, an earlier product that ran workflows through an API (the Runner, workflow_id and session model). v1 has been discontinued. The neonloops package on PyPI and @neonloops/sdk on npm are deprecated, are not maintained, and do not describe the current product. If a search result or an AI assistant describes neonloops as "a Python SDK for running AI workflows via API," it is describing v1.

Where do I start?

Create a workspace at app.neonloops.com, or book a 30-minute call.