Installation
This page is the complete install guide: official packages, the CLI one-liner, and a source build of the gateway + WebUI.
Related:
- First session in the terminal: CLI quickstart
- One-liner only: CLI install
- Product map: Capabilities
- Site: navin.live/download · navin.live/en/docs
After any official install you should have two commands:
| Command | Role |
|---|---|
navin-cli | Terminal AGI in the current folder (this is the product CLI) |
navin | Desktop / workbench, doctor, status, gateway, navin . |
Do not use navin tui or navin agent as the entry point.
1. Desktop packages (recommended)
Download from navin.live/download.
| Platform | Files | Notes |
|---|---|---|
| Windows | .exe setup or .msi | Signed by Navinspire. SmartScreen may still show: More info → Run anyway |
| macOS | .dmg (arm64 or x64) | Drag Navin into Applications. If Gatekeeper blocks: right-click → Open |
| Linux | .AppImage, .deb, .rpm, .pkg.tar.zst | AppImage may need libfuse2 on older distros |
User data stays in ~/.navin (or %USERPROFILE%\.navin on Windows) across upgrades: config, workspaces, memory, projects.
Windows SmartScreen ("Windows protected your PC") can appear on a correctly signed setup. Microsoft checks the publisher and the reputation of that exact file hash. Navinspire as signer does not clear it on the first downloads. An EV certificate no longer bypasses it. Keep More info → Run anyway until reputation builds, or install from the Microsoft Store when that listing exists. To check the file you actually downloaded:
Get-AuthenticodeSignature ".\Navin-Desktop-2.0.1-windows-x64-setup.exe" |
Format-List Status, StatusMessage, SignerCertificate
Status should be Valid. That confirms the signature, not SmartScreen. The same check, including navin.exe / Navin.exe / NavinUpdater.exe inside the installer, is packaging/windows/verify-windows-signatures.ps1.
If navin or navin-cli is missing from PATH after a DMG or portable install:
navin install-cli
navin install-cli --force
Then:
cd your-project
navin-cli
navin .
navin --version
navin doctor
2. Official CLI one-liner
The script reads the official releases.json. It prefers a CLI archive when published, otherwise it extracts the desktop package. It writes navin and navin-cli to the user PATH. No sudo on Linux. Default prefix: ~/.local.
Linux, macOS, WSL:
curl https://navin.live/install -fsS | bash
Windows PowerShell:
irm 'https://navin.live/install?win32=true' | iex
cd your-project
navin-cli
navin --version
navin doctor
navin status
Linux arm64: use navin.live/download until a CLI tarball is listed.
Environment overrides:
| Variable | Meaning |
|---|---|
NAVIN_DOWNLOAD_BASE | Alternate download root |
NAVIN_SITE | Site used by the script (default https://navin.live) |
NAVIN_PREFIX | Install prefix (default ~/.local) |
Local site (developers)
When the marketing site runs on this repo (localhost:3100) and you have prepared local archives:
make local-releases
curl http://localhost:3100/install -fsS | bash
/install then serves files from this repository and does not hit production S3.
3. First configuration
In navin-cli, Ctrl+G → Providers: add a key (OpenRouter, OpenAI, Anthropic, …) or a local apiBase (Ollama http://127.0.0.1:11434/v1). Then Models: add a configuration and set it active.
Or:
navin onboard
Env keys (OPENROUTER_API_KEY, …) are honored and are not written to config.json.
This tree is BYOK only. Add keys in Ctrl+G → Providers. There is no Navin managed provider and no navin.live account on main.
Config file: ~/.navin/config.json. Same file for navin-cli and the desktop.
4. Build from source (gateway + WebUI)
Use this when you clone the repository and want to run or change Navin locally.
make install installs missing system packages when it can (apt on Debian/Ubuntu, dnf/yum on Fedora/RHEL, pacman on Arch, zypper, Homebrew on macOS), then the Python backend and the WebUI.
Prerequisites
If you skip system install (NAVIN_SKIP_SYSTEM=1 or --no-system):
- Python 3.11 or newer
- Make
- Git
- Node.js 18+ and npm
- Linux / macOS: rustup if you need the native sandbox (
make native)
Two commands
git clone https://github.com/navinspire-ai/navin-agi.git
cd navin-agi
make install
make start
Or the same with scripts:
sh scripts/install.sh
sh scripts/start.sh
One shot (install + start):
sh scripts/start.sh --install
This is dev:
- Vite hot reload:
http://localhost:5173/ - the gateway (health + API). The same gateway serves the production WebUI from
navin/web/distathttp://localhost:8765/(orchannels.websocket.port)
make stop
# or: sh scripts/stop.sh
Front / backend
Make and scripts accept the same scopes:
| Command | What it does |
|---|---|
make install / sh scripts/install.sh | System + backend + WebUI |
make install backend | Backend only (.venv) |
make install front | WebUI only (npm ci) |
make start / sh scripts/start.sh | DEV: gateway + Vite http://localhost:5173/ |
make start-prod / sh scripts/start.sh --prod | PROD: build if needed + gateway (no Vite) |
make build / sh scripts/build.sh | Front (navin/web/dist) + backend (pip editable) |
make build front | Vite production bundle only |
make build backend | Refresh .venv (pip install -e .) |
make start backend | Gateway only (serves the build on :8765) |
make start front | Vite only |
make start-fg / sh scripts/start.sh --fg | Gateway in the foreground |
make stop / restart / status | Same scopes: front or backend |
make start backend
make start front
make restart
make status
make logs
Native helpers (Linux / macOS), after make install:
make native
Vite proxies /api, /webui, /auth and the WebSocket to the gateway. If the gateway is not on 8765:
NAVIN_API_URL=http://127.0.0.1:8766 npm run dev
Docker (this repo)
No third-party Navin image. Build from https://github.com/navinspire-ai/navin-agi (this tree):
docker compose up navin-gateway
The image installs the navin CLI and serves the bundled WebUI. Gateway health defaults to 18790, WebUI to 8765. Config: ~/.navin mounted into the container.
Use the source CLI
From the navin-agi clone, after install, you can launch the CLI with the venv binary. No need to activate the venv:
cd /path/to/navin-agi
.venv/bin/navin-cli
Windows:
cd \path\to\navin-agi
.venv\Scripts\navin-cli
Or activate the venv, then run it from any project folder:
source .venv/bin/activate # Windows: .venv\Scripts\activate
cd /path/to/your-project
navin-cli
navin doctor
Ports
| Role | Default | Config |
|---|---|---|
| WebUI + WebSocket | 8765 | channels.websocket.port |
| Gateway health | 18790 | gateway.port |
| OpenAI-compatible API | 8900 | api.port (navin serve) |
| Vite dev | 5173 | WebUI Makefile |
navin ports
navin ports check
See Ports.
Always-on gateway
navin gateway --background
navin gateway status
navin gateway logs
navin gateway install-service
Long-running agent · Deploy the gateway
5. Uninstall
- Desktop: use the OS uninstaller.
- CLI prefix: remove
~/.local/bin/navin,~/.local/bin/navin-cliand~/.local/share/navin(or yourNAVIN_PREFIX). - User data is not deleted automatically. Remove
~/.navinonly if you want to wipe config, chats and workspaces.
navin cache --clear
Does not delete chats, config, or workspaces.
6. Troubleshooting
navin doctor
navin --version
navin ports check
| Symptom | What to try |
|---|---|
navin-cli: command not found | New terminal, or navin install-cli, or export PATH="$HOME/.local/bin:$PATH" |
| Gateway will not start | navin ports check, then make logs |
| Vite cannot reach the API | Confirm gateway port in ~/.navin/config.json, set NAVIN_API_URL |
| No model replies | Ctrl+G → Providers / Models, or navin doctor |
More: CLI troubleshooting · Start without a technical background