NeoRecall documentation
NeoRecall is a self-hosted audio memory service. It records from web, desktop, and wearable clients, keeps its durable database and search index on the server you install, and sends inference to transcription and language-model services you configure separately. It never downloads or runs a speech-recognition or language model; the only inference it does itself is detecting speech and telling voices apart, which no such service can hand back.
NeoRecall is beta software. Install it on a machine you administer and read the privacy guide before recording anyone else.
🚀 Start here
bash <(curl -fsSL https://raw.githubusercontent.com/NeoLabs-Systems/NeoRecall/main/install.sh)
The installer clones NeoRecall from GitHub, links the global neorecall
command to that checkout, and runs the guided install. Desktop app users can do
the same from Set up NeoRecall on this computer without a terminal.
Open http://localhost:4500 and /app/ when setup finishes. Continue with:
Configure both inference providers before recording; the server intentionally has no bundled transcription or generation fallback.
🧭 User guide
| Guide | Use it for |
|---|---|
| Installation | Host requirements, repository installer, Docker, service commands |
| Recording | Web, macOS, and Windows capture behavior and offline sync |
| Configuration | Environment variable reference |
| Privacy and consent | What the server stores and how to record responsibly |
| Troubleshooting | Diagnosing setup, sync, and startup issues |
🛠️ Developer guide
The developer guide explains the implementation rather than the product setup. Start with Architecture.
Contributors must also follow GUIDELINES.md, AGENTS.md, and CONTRIBUTING.md.