Nelly – Build Your Own Team of AI Agents

TL;DR

TL;DR: Nelly is a cross-platform desktop app that lets you build, test, and run teams of AI agents using natural language — no code required. In open beta on macOS, Windows, and Linux.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: nelly.is — verified, returned HTTP 200 on 2026-06-30
• HN launch thread: news.ycombinator.com/item?id=43914754 — 54 points
• Company: Pine Enterprises AB (Sweden)
• License: Proprietary (closed source, no public license found)
• Source last checked: 2026-06-30

What Is Nelly?

Nelly is a desktop AI agent platform that lets you build, test, distribute, and use custom AI agents without writing code. Instead of prompting a single chat interface, you compose a team of specialized agents that can delegate tasks to each other and work together on complex workflows.

The product is structured around three core concepts:

• Studio — build agents using natural language instructions. Describe what you want the agent to do, format it with headings and lists, and assign it tools.
• Chat — have natural conversations with your agents. Threads keep conversations organized and improve agent context over time.
• Marketplace — share and monetize agents (coming soon).

Nelly also supports Sub-Agents: break a complex task into smaller problems and delegate them to specialized agents, so a single team can handle multi-step workflows autonomously.

Platform Support

Nelly is currently in open beta across three desktop platforms:

| Platform | Download | |---|---| | macOS (Apple Silicon) | .dmg via nelly.is/download | | macOS (Intel) | .dmg via nelly.is/download | | Windows 10+ | .exe via nelly.is/download | | Linux 64-bit | .AppImage/.deb via nelly.is/download |

The company behind Nelly is Pine Enterprises AB, based in Sweden.

Setup Workflow

Step 1: Download and Install

Visit nelly.is/download and pick the installer for your OS. Nelly is currently in beta — if demand is high you may be placed on a waitlist, but downloads are available for all three platforms.

Step 2: Build Your First Agent in Studio

Open Nelly and navigate to Studio. Describe your agent in natural language:

You are a research assistant. Your job is to find the most
relevant articles on a given topic, summarize the key points,
and present them in bullet format.

Format your instructions using headings, bullet lists, and numbered steps — Nelly reads these as structural cues for how your agent should behave.

Step 3: Assign Tools

In Studio, attach tools to your agent. Available tool categories include:

• Browser — give your agent web access
• Database — connect to structured data sources
• More tool integrations are listed as “coming soon” on the product site

Step 4: Test in Chat

Switch to Chat and start a conversation with your agent. Use threads to organize different projects. The activity panel shows what tools your agent used, browser screenshots, and step-by-step progress.

Step 5: Use Sub-Agents for Complex Tasks

For multi-step workflows, open the Sub-Agents panel and define specialized agents. For example, a “product research” team might include:

• A web search agent to gather competitor info
• A summarizer agent to condense findings
• A writer agent to format the final report

The orchestrator agent delegates to sub-agents and aggregates their outputs.

Deeper Analysis

What Makes Nelly Different

Most AI agent platforms focus on a single agent interacting via chat. Nelly’s team model mirrors how you would organize a real workplace — specialists handle specific domains, and a lead agent coordinates. The Sub-Agent feature is the key differentiator: it lets one agent break down a problem and farm out pieces without human intervention in the loop.

The platform is also genuinely cross-platform from day one (macOS, Windows, Linux), which is rare for desktop AI agent tools that often start macOS-only.

Limitations

• Closed source — no public codebase to audit. Nelly is a proprietary product from Pine Enterprises AB.
• Beta stage — the product is still in active development. Features like the Marketplace are listed as “coming soon.”
• No self-hosted option — agents run on Nelly’s infrastructure. There is no mention of an on-premises or self-hosted deployment model.
• Custom LLM providers are a paid feature — if you want to connect your own model providers rather than using Nelly’s defaults, that requires a paid plan.

Practical Evaluation Checklist

• [ ] Download the beta for your OS from nelly.is/download
• [ ] Build an agent in Studio using natural language
• [ ] Attach a tool (browser or database) and test it in Chat
• [ ] Try a multi-step task with Sub-Agents
• [ ] Organize a project using threads in Chat
• [ ] Review the Activity panel to see tool usage and screenshots

Security Notes

• Nelly agents have access to tools like browsers and databases — grant access mindfully, especially when connecting to live data sources.
• There is no mention of data retention policies, agent audit logs, or sandboxing in the publicly available documentation.
• As a proprietary closed-source product, the backend cannot be independently audited.

FAQ

Q: Is Nelly open source? A: No. Nelly is a closed-source proprietary product from Pine Enterprises AB. There is no public source repository.

Q: Can I connect my own LLM provider? A: Yes, but only on a paid plan. Custom LLM provider integration is listed as a paid feature.

Q: Is there a self-hosted option? A: No. Nelly runs as a desktop application with a cloud backend. There is no on-premises deployment option documented.

Q: What operating systems are supported? A: macOS (Apple Silicon and Intel), Windows 10+, and Linux 64-bit. All currently in open beta.

Q: Is there a waitlist? A: Possibly, depending on beta demand. The download page notes you might be placed on a waitlist during the beta period.

Q: How does Nelly compare to building agents with LangChain or CrewAI? A: Nelly targets non-technical users — no code, natural language only. LangChain and CrewAI are code-first developer frameworks. If you want to prototype agent workflows without writing Python, Nelly is faster; if you need full customization and self-hosting, code-first frameworks offer more control.

Conclusion

Nelly fills a gap for non-technical users who want the power of multi-agent AI teams without touching code. The natural-language Studio, Sub-Agent delegation, and cross-platform availability make it distinctive among AI agent platforms. The closed-source, cloud-dependent model is the main limitation — there is no self-hosted path and no public codebase to audit.

If you want to explore team-based AI agents and prefer a polished desktop app over writing orchestration code, Nelly is worth a try during its open beta. Watch the Marketplace feature closely if you are interested in sharing or monetizing agents.

Try it: nelly.is/download