SovereignAI

SovereignAI is a self-hosted private AI command center that runs entirely on the user's own hardware with zero runtime dependencies, providing a chat interface, durable memory, and a personal knowledge base without any cloud or account requirement. The platform supports local open-weight models via Ollama and FreeToken, and every memory carries a receipt documenting its origin, while all data can be exported into a single checksummed, optionally encrypted file for verified exit.

Traction –/10

HQ unknown

Artificial Intelligence
AI Agents
Software Only

Updated 5 days ago

Funding

Funding rounds are not available yet.

Founders

Founder details are not available yet.

Product

Problem

Users' personal data is fragmented across dozens of services, each holding a partial profile that can be monetized, while AI assistants accumulate transcripts of user thoughts in accounts users do not control. The growing dependence on cloud-based AI creates a "moat" of accumulated memory that makes leaving a service increasingly costly, and no single entity holds the complete picture of a person's life.

Solution

SovereignAI is a private AI command center that runs as a single program on the user's own laptop, homelab, or server, with a web interface, a chat that knows the user, durable memory, and a knowledge base built from personal documents. The software is written with zero runtime dependencies, making the code auditable, and works fully offline with no account or cloud requirement. Every memory carries a receipt recording how it entered, which conversation it came from, and which model wrote it, while deletion zeroes bytes on disk. The platform supports local open-weight models through Ollama and FreeToken for frontier-class sparse models, with remote model use disclosed on-screen at the moment of use. Before any request goes to a remote provider, users see a "customs declaration" showing exactly what would leave their machine. All data can be exported into one checksummed, optionally encrypted file in a documented format, providing a verified exit path.

Target Audience

Primary users are individuals and organizations who want full ownership and control over their AI assistant's data, models, and infrastructure, including privacy-conscious professionals, developers, and homelab operators running AI on their own hardware.

Features

Zero-dependency core written with only a modern Node runtime's standard library, enabling literal code auditability
Memory system that records origin, source conversation, and authoring model for every fact, with disk-zeroing deletion
On-machine parsing of PDFs, Word documents, notes, and imported chat history from ChatGPT and Claude without uploading content
Support for local models via Ollama, FreeToken for sparse mixture-of-experts models on consumer GPUs, and any OpenAI-compatible server
Pre-request "customs declaration" showing system prompt, injected memories, retrieved document excerpts, and prior messages with sizes before remote calls
One-file checksummed export with per-table SHA-256 manifest, optional AES-256-GCM encryption, and a separate verify command
MCP server integration for Claude Desktop, Claude Code, Codex CLI, Cursor, Windsurf, and Gemini CLI, plus browser and IDE extensions
Fair-source license with MIT licensing after two years