Hi, my name is Timothy and I've been working on Nelly for the last few months.
Nelly is an AI agent/assistent platform where you can build, use and soon share AI agents. You create the agents by simply writing instructions in natural language, and you can break down these instructions into smaller parts as sub-agents. You then use these agents with Nelly's chat interface to perform tasks for you.
Nelly is a desktop app, currently for Mac but soon also for Windows and Linux, that you install on your computer. The agents and their data are stored on your computer but the LLM used for decision making, etc. is running in the cloud (currently Claude 3.7 Sonnet via OpenRouter). With a paid subscription you can connect Nelly directly to your own LLM providers.
You will soon be able to both share and explore third-party tools and agents on the Marketplace. Built-in tools at the moment include a browser, a calculator, and a database.
Read about how I use Nelly to pay out my salary, from calculating the net salary, making the bank transfers, and declaring it to the tax authorities, to recording it in my accounting software:
X: https://x.com/timothyej/status/1...
LinkedIn: https://www.linkedin.com/posts/t...
This is just the beginning, the next step is to add MCP support, then launch the Marketplace with both agents and tools, and improving the platform with file support and more.
I'm looking forward to see what you build!
Installed Nelly last night and spent half an hour just roaming around the interface; even at first glance, the no-code builder feels intuitive—slotting in sub-agents and toolkits is nearly drag-and-drop.
One idea that could speed up onboarding: a public gallery of ready-made agents (think SEO audits, inbox triage, churn-risk pingers). Starting from a template would slash time-to-value and jump-start your upcoming marketplace.
Minor blocker for me: I split my day between Mac and Windows. Any ETA on Windows support or a browser-based option? Once that’s live, I can roll Nelly out across my whole team. 😬
@gianmaria_caltagirone Great feedback! 🔥 Many agents/sub-agents have similar instructions and being able to start from a template would make a lot of sense and save a lot of boilerplate typing.
Windows-support is not far away, probably within 2 months or so 🤞 Sign-up for the waitlist to get notified: https://nelly.is/windows-waitlist
Until then, I would love to learn more about your needs and use-cases, feel fee to contact me directly at timothy@nelly.is or on any of the following platforms:
@timothyej Thank you very much for your answer! I’ve just joined the waitlist for Windows version.
Let me try Nelly a little bit and I’ll be glad to give you more feedback. 🙏
I just tried building a simple agent to fill out a form using the provided data, and it worked really well- 3 out of 4 forms were submitted successfully, so there’s still a bit to improve, but overall it’s promising. Earlier today, I accomplished the same task using Automatio.io, but the key difference was setup time: Automatio took me about 10 minutes (and I had to click a lot!), while Nelly handled it in just 1 minute. I’m definitely planning to push it further with more complex scenarios. Combining NLP and RPA really feels like the future of automation
@przemyslaw_szustak Nice, that's awesome, thanks for testing! 🔥 Do you know what happened with the failed attempt? Feel free to provide feedback directly to me at timothy@nelly.is or on any of the following platforms:
Tip: Move the Form Builder to a sub-agent and then have the main agent call the sub-agent with each number. That should increase reliability and speed (but it will consume more message credits).
Keep the app updated - new improvements are rolled out almost everyday! 🚀
Nelly
@timothyej Can Nelly’s agents integrate with custom APIs or external services beyond the built-in tools?
Nelly
Installed Nelly last night and spent half an hour just roaming around the interface; even at first glance, the no-code builder feels intuitive—slotting in sub-agents and toolkits is nearly drag-and-drop.
One idea that could speed up onboarding: a public gallery of ready-made agents (think SEO audits, inbox triage, churn-risk pingers). Starting from a template would slash time-to-value and jump-start your upcoming marketplace.
Minor blocker for me: I split my day between Mac and Windows. Any ETA on Windows support or a browser-based option? Once that’s live, I can roll Nelly out across my whole team. 😬
In the meantime, great work! 🙏
Nelly
@gianmaria_caltagirone Great feedback! 🔥 Many agents/sub-agents have similar instructions and being able to start from a template would make a lot of sense and save a lot of boilerplate typing.
Windows-support is not far away, probably within 2 months or so 🤞 Sign-up for the waitlist to get notified: https://nelly.is/windows-waitlist
Until then, I would love to learn more about your needs and use-cases, feel fee to contact me directly at timothy@nelly.is or on any of the following platforms:
X: https://x.com/nellydotis
Discord: https://discord.gg/RGVmdya2Up
Reddit: https://www.reddit.com/r/nellyintelligence
Thanks for the feedback!
Tidio
I just tried building a simple agent to fill out a form using the provided data, and it worked really well- 3 out of 4 forms were submitted successfully, so there’s still a bit to improve, but overall it’s promising. Earlier today, I accomplished the same task using Automatio.io, but the key difference was setup time: Automatio took me about 10 minutes (and I had to click a lot!), while Nelly handled it in just 1 minute. I’m definitely planning to push it further with more complex scenarios. Combining NLP and RPA really feels like the future of automation
Nelly
@przemyslaw_szustak Nice, that's awesome, thanks for testing! 🔥 Do you know what happened with the failed attempt? Feel free to provide feedback directly to me at timothy@nelly.is or on any of the following platforms:
Discord: https://discord.gg/RGVmdya2Up
Reddit: https://www.reddit.com/r/nellyintelligence
X: https://x.com/nellydotis
Tip: Move the Form Builder to a sub-agent and then have the main agent call the sub-agent with each number. That should increase reliability and speed (but it will consume more message credits).
Keep the app updated - new improvements are rolled out almost everyday! 🚀