Sovereign AI that runs on your machine, remembers in chains you own, and answers to no one but you. Local-first, chain-backed memory, vault-sealed secrets, zero vendor lock-in.
Memphis isn't for everyone. It's for operators, developers, and teams who refuse to put their AI infrastructure in someone else's hands.
Local runtime on your own hardware. No cloud dependencies, no vendor lock-in. Real health signals, structured logs, CLI you can script.
Chain-backed memory you can audit. Vault for secrets. RAG-ready. Provider-agnostic — Ollama, OpenAI, Anthropic, local models all work together.
Self-hosted, sovereign infrastructure. No data leaves your network. Designed for teams that need compliance, privacy, or just不想 довіряти cloud.
Memphis running on WSL2 (Ubuntu) on a Windows machine. MiniMax-M2.7 model, full terminal UI, chain-backed memory, vault integration.
WSL2 + Ubuntu + Memphis runtime + MiniMax-M2.7 · Memphis health check in terminal
For the basic user the most important question is not "does it sound smart", but "what do I actually get after installation". The answer is simple: local runtime, memory, vault, cockpit and the ability to work with providers without locking yourself into one vendor.
Linux, macOS or WSL2. Memphis runs locally under your control.
Journal, decisions and system continuity live in append-only chain, not random prompt log.
Keys and tokens are encrypted, not scattered across envs and scripts.
TUI, HTTP API and agent integrations give real system access, not just README.
Linux, macOS or WSL2. Installs Node 22, Rust stable, Ollama, clones the repo, builds everything, and links the memphis CLI globally.
One command, full stack installed automatically.
Add --with-init to chain first-run (vault passphrase, identity, provider enrollment) in the same session.
Verify your runtime is healthy and ready.
Ends with a running operator instead of a linked CLI.
If after this page you already know Memphis is for you, the next move should be simple. You don't have to click around blindly.