Skip to contentDiscuss a freight pilot →

AI back office for freight brokers.
Agents do the paperwork. A person checks the work.

← All posts
July 15, 2026 · Shiva Saini

Why I Built NanoCrew

Today's AI agents generally fall into one of two categories:

  • General-purpose agents built to handle almost anything reasonably well.
  • Specialized agents built for one specific domain, like a coding assistant.

But what if you don't want either?

What if your business has a unique workflow that no existing AI agent is designed to solve?

Today, you have two choices:

  1. Use a general-purpose agent and hope it adapts to your workflow.
  2. Hire a team to build a custom AI agent for your business.

The second option is expensive, time-consuming, and repetitive. Every new business problem usually requires building another specialized agent from scratch. If you need 10 different agents, you're likely paying to build 10 different systems — with no guarantee they'll actually solve your problems the way you expect.

That's the problem I built NanoCrew to solve.

Give each agent a role

NanoCrew is an open-source platform for running a crew of AI agents. You create an agent from a template or from scratch. Each agent gets a workspace, an identity, long-term memory and a set of tools, and you manage the crew from a web dashboard.

The current release is v0.1 and early. Setup requires Docker, Node.js and your own model-provider account. It is not a hosted service that works without configuration.

Keep the work visible and bounded

Every agent cycle runs in a Docker sandbox. Sensitive actions, such as sending email, go to an approvals inbox so a person can review them. Per-cycle logs and run history help you inspect what happened.

Your agents also don't work in isolation.

Each agent specializes in a specific responsibility, and they can hand off work to one another whenever necessary. Together, they function like a real team, collaborating to solve complex business problems that would be difficult for a single agent to handle alone.

Long-term memory can carry context between runs; it is not a guarantee that an agent will improve or complete a task correctly.

Work the way that fits you, not the other way around

NanoCrew also gives you complete flexibility in how your agents operate. You can:

  • Run them on demand.
  • Schedule them to work autonomously.
  • Operate them directly from WhatsApp.
  • Integrate them into your existing applications, products, or workflows through APIs.

The vision

My vision for NanoCrew is simple:

AI shouldn't force businesses to adapt to generic agents. Businesses should be able to create AI agents that adapt to them.

Instead of buying one AI agent after another — or hiring engineers every time your business evolves — NanoCrew lets you build, deploy, and configure a crew of specialized AI agents that can work together on your workflows.

That's why I built NanoCrew.


Want to see it for yourself? Explore NanoCrew in Open source or read how choosing a self-hosted workflow.