Navin
All documentation

How to Keep a Self-Hosted AI Agent with Navin

This guide frames Navin as a self-hosted AI agent that stays on your computer. The desktop app keeps chats, projects, memory, and provider keys local unless you choose otherwise.

What you get

  • Navin running as a local desktop app under your control
  • model providers configured in Settings
  • optional chat apps and automations while the app stays open
  • data that lives in your projects on disk

When to use this

Use this path when you want local ownership of the agent process, project files, memory files, and provider keys - without sending everyday work to a hosted chat product.

Keep Navin on your machine

  1. Install Navin from navin.live/download.
  2. Open the app and complete the setup wizard.
  3. Add a provider under Settings → Providers (your own API key or a local model server such as Ollama).
  4. Set an active model under Settings → Models.
  5. Open a project folder you control and chat there.

Chats, files, and memory stay on your computer. An optional navin.live account syncs plan, license, and device activations only - see Privacy in ../start-without-technical-background.md.

Stay available for channels and automations

Leave Navin running when you need:

  • Telegram, Discord, Slack, and other channels (Settings → Channels)
  • scheduled automations and heartbeat checks
  • long-running goals in chat

You do not deploy a separate server for everyday personal use. The desktop app is the local runtime.

Production notes

  • Use one project per trust boundary.
  • Prefer Settings for secrets; avoid pasting keys into shared documents.
  • Review tool access (Settings) before inviting teammates through chat apps.
  • For separate clients or bots, use separate projects and channel tokens - see ../multiple-instances.md.

Security notes

  • Keep Navin local unless you intentionally share access on a LAN.
  • Prefer pairing for DM-capable chat apps; keep allowlists strict.
  • Enable project / workspace restriction before broad file or shell tools.
  • On Linux, prefer the shell sandbox when Settings offer it.

Troubleshooting

  • Provider errors: fix Settings → Providers / Models, then retry a short chat.
  • Channels idle: confirm the channel is enabled and Navin is still open.
  • Wrong memory or files: confirm the active project in the project picker.

Related docs