/ terminal client for nexora

The whole platform,
in your terminal.

Chat with your agents, watch tasks run live, and let them work on your own machine. One quick install, and it connects to any Nexora instance you run.

GitHub·star it
dev@acme: ~/workprod
$ nexora login --url https://nexora.acme.internal --name prod
✓ Logged in as dev@acme.internal. Saved instance "prod".
$ nexora
 
◆ Nexora●chatsessionsagentsboardtaskssettings
dev@acmeDeploy the staging build and tell me if the health checks pass.
 
-activity
⚡ ✓ shell_run: kubectl apply -f k8s/staging/ (12 lines · ctrl+l)
⚡ ✓ http_get: https://staging.acme.internal/healthz 200 OK
 
▸ ◆ Verify staging rollout ✓
▸ watch rollout status deploy/web
▸ probe /healthz across all pods
 
● Local Operator
Staging is live. All 6 pods are ready and /healthz returns 200
across the board. Rollout finished in 41s.
Want me to promote this build to production?
connect · Connect to any instance

Installed in one command.

Paste one line and you are done. Nothing else to install, nothing left running in the background. The command below sets up the nexora command; run it again any time to update.

linux / macos

windows · powershell

or build from source

$ make build-all # linux · darwin · windows, in a Go container

Then point it at your instance.

Nexora CLI connects to your own Nexora or NexoraCloud server, the same one the browser uses. Sign in whichever way suits you.

$ nexora login --url https://…

Email and password, with TOTP when 2FA is on.

tokens live in config.toml (mode 0600), auto-refresh on 401, and never get committed.

/ feature coverage

Everything the web UI does.
Now keyboard-driven.

Every feature from the web app is here too. Thirteen screens, each a keystroke away, all showing the same live data.

nexora · chat
The NexoraCLI chat screen: a streaming reply from the Local Operator agent with a tool call and the agent hierarchy panel.

tab · move between screens

01 / 13

● chatscreen

Streaming chat, slash commands, thinking blocks.

nexora · chat
┌ dev@acme
Summarize today's failed deploys.
─ activity ────
⚡ ✓ http_get: /api/deploys?status=failed
● Local Operator
Two failed today. Both on staging,
image pull backoff. Want the logs?
reach it·tab to cycle, or ctrl+k then type chat

Agents that work on your machine.

Let an agent read files and run commands right where you are working, not just on the server. Your data, your keys, and your access stay on your own computer. You are always in control: nothing runs until you allow it.

  1. 01

    Turn it on

    /local · /yolo

    One toggle switches it on, and a LOCAL badge shows in the footer. It is off until you enable it, and your machine, not the server, is always in charge.

  2. 02

    Point it at a folder

    /cd ~/work/acme

    Tell it which folder to work in, and change it any time. "List my desktop" really lists your desktop, because the work happens where you are.

  3. 03

    Approve what matters

    run migrate.sh? [y/n]

    Safe, read-only steps run on their own. Anything that changes your files asks first, so nothing happens by surprise.

  4. 04

    See everything it does

    ⚡ ✓ shell_run

    Every command shows up live as it runs, with its full output one keypress away. Nothing is hidden, nothing is cut short.

local executionLOCALYOLO off

$ agent wants to run:

⚡ shell_run: ./deploy/migrate.sh --prod

→ approve this write?[y]/[n]

⚡ ✓ shell_run (8 lines · ctrl+l)

applied 3 migrations in 1.2s

⚡ ✓ file_read ./VERSION auto

read-only and safe shell auto-run. writes and arbitrary shell prompt for consent. denials return an error so the turn never hangs.

Never reach for the mouse.

Every screen, panel, and action is a keystroke. Hit ctrl+k anywhere for the command palette, or type / to fire an in-chat command with autocomplete.

ctrl+k · command palette

> switch to

  • agents · manage agents
  • providers · llm providers
  • board · kanban
switch screentab
command palettectrl+k
send / openenter
toggle side panelctrl+b
cycle panelsctrl+o
sub-chat navigatorctrl+g
copy last replyctrl+y
expand outputctrl+l
slash commands/
back / cancelesc

in-chat slash commands

/new/agent/model/chain/local/yolo/cd/usage/copy/help

/ self-hosted by design

Your data never leaves
your network.

Nexora CLI talks only to the server you run, the same way the browser does. No tracking, no company in the middle. It is something hosted AI tools simply cannot offer, and with self-hosted Nexora you get it by default.

REST
<url>/api/*

Authorization: Bearer <jwt | nxr_key>, transparent refresh on 401

WS
<url>/ws/chat/{id}

chunk · tool_call · sub_agent · task_updated frames, live

SWITCH
nexora instance use prod

hold many instances, switch the active one without re-auth

~9.5 MB
quick download
0
extra setup
3
operating systems
13
screens, same as web

One instance, three clients.

Run the platform however you work. The CLI is one of three first-party clients that all talk to the same self-hosted instance.

Webbrowser

The full console for building agents and running the platform.

open →
CLIterminal

This one. Everything the web app does, in your terminal.

you are here
MobileiOS · Android

Scan a code once, then chat with your agents on the go.

open →

Get Nexora CLI.

One command to install. Then run nexora and connect to your instance.

GitHub·star itread the docs→

one line, any platform