NeuralCleave

NeuralCleave is an open-source personal AI assistant that brings every AI model, conversation, memory, and workflow together in one private, extensible intelligence layer. Rather than tying you to a single app, it runs one assistant across 32 channels — Telegram, WhatsApp, Discord, Slack and 29 more — and 13 LLM providers including Anthropic, OpenAI, DeepSeek and Ollama, so you reach the same assistant wherever you already chat. Its 3-tier memory system uses Redis for hot session data, Qdrant for vector semantic recall, and SQLite for long-term storage, with per-agent namespace isolation, compaction and an archiver, and it works fully offline. Task-aware model routing automatically sends 10 task types to the optimal provider — complex reasoning to Claude Opus, code to DeepSeek, cheap inference to Ollama — and a privacy mode can force all calls to run locally. NeuralCleave is voice-ready with Whisper speech-to-text, 3-tier text-to-speech and wake-word support across platforms, and it ships a plugin SDK with typed abstract base classes, entry-points, hot-reload and a Hub marketplace. The current v2.1.5 release advertises 41 REST endpoints and thousands of passing tests. It fits developers and privacy-minded power users who want a self-hosted, model-agnostic assistant they fully control rather than a closed cloud chatbot.

Best for developers and privacy-minded power users who want a self-hosted, model-agnostic AI assistant across many chat channels.