No reviews yetBe the first to leave a review for NeuralCleave
Maker
📌
Hey Product Hunt 👋 I'm Amit, the solo developer behind NeuralCleave.
I built this out of a personal frustration. I was juggling
ChatGPT in one tab, Claude in another, a Telegram bot I'd
set up months ago, and Ollama running locally — none of them
sharing memory, none talking to each other.
I wanted ONE agent. On MY machine. That works across every
platform I actually use.
A few things I'm particularly proud of:
🧠 The ReflectionEngine — every response is automatically
scored across Relevance, Completeness, Accuracy and Tone
(0–100). If it fails the threshold it gets regenerated. You
only ever see passing responses.
🔒 Privacy mode — forces all inference through Ollama. Zero
data leaves your machine. Zero API calls to anyone.
📡 32 channels — from Telegram and Discord to Nostr,
Tlon/Urbit and iMessage. Same agent, same memory, everywhere.
It's been about a year of building. 5,064 tests. v2.1.0
shipped last week.
pip install neuralcleave to try it — runs in under a minute.
Happy to answer anything — what would you want to connect
your AI agent to first? 🙏
Report
Maker
🚀 Update — v2.1.4 shipped on July 27!
Fixed the Windows "Gateway online" bug that was blocking some users on
first install — root cause was a Tauri v2.11+ CORS origin mismatch on
Windows 11. The app now also self-heals automatically after a crash
(no reboot needed).
Desktop installers for Windows, macOS, and Linux are live:
Been using NeuralCleave as my actual daily driver for the past few weeks. Three things broke or annoyed me enough to fix them.
🔧 Terminal CLI routing — fixed
The embedded terminal lets you run `neuralcleave` commands inside the desktop app. In v2.1.4 these were silently broken — the command ran, prompt returned, nothing looked wrong, but the gateway response was being discarded before it reached the output. Fixed.
Commands that now work: `neuralcleave status`, `channels list`, `memory search`, `chat`
🧠 Multi-phase thinking indicator
Instead of a static spinner, the chat now cycles through:
Same pattern as Claude's UI. A 6-second wait with a spinner feels broken. The same 6 seconds with status messages feels intentional.
↔️ Dynamic chat width + toolbar redesign
Chat area now expands when the sidebar collapses, contracts when it opens — smooth CSS transition. History and New Chat buttons redesigned into a proper grouped pill control.
🚀 Update — v2.1.4 shipped on July 27!
Fixed the Windows "Gateway online" bug that was blocking some users on
first install — root cause was a Tauri v2.11+ CORS origin mismatch on
Windows 11. The app now also self-heals automatically after a crash
(no reboot needed).
Desktop installers for Windows, macOS, and Linux are live:
→ https://github.com/TheAmitChandra/NeuralCleave/releases/tag/v2.1.4
Thanks for all the support today 🙏
Hey everyone 👋 Maker here with a v2.1.5 update —
Been using NeuralCleave as my actual daily driver for the past few weeks. Three things broke or annoyed me enough to fix them.
🔧 Terminal CLI routing — fixed
The embedded terminal lets you run `neuralcleave` commands inside the desktop app. In v2.1.4 these were silently broken — the command ran, prompt returned, nothing looked wrong, but the gateway response was being discarded before it reached the output. Fixed.
Commands that now work: `neuralcleave status`, `channels list`, `memory search`, `chat`
🧠 Multi-phase thinking indicator
Instead of a static spinner, the chat now cycles through:
Thinking → Searching memory → Analyzing context → Reviewing context → Crafting response
Same pattern as Claude's UI. A 6-second wait with a spinner feels broken. The same 6 seconds with status messages feels intentional.
↔️ Dynamic chat width + toolbar redesign
Chat area now expands when the sidebar collapses, contracts when it opens — smooth CSS transition. History and New Chat buttons redesigned into a proper grouped pill control.
---
5,076 tests passing. PyPI updated.
- ⬇️ neuralcleave.com
- 🐍 `pip install --upgrade neuralcleave`