Run a team of coding agents
on your GPU server

Claude Code, Codex, and 40+ coding agents, running where your GPUs are.
Start them at your desk, steer them from your phone.

Get Vicoa for free

Works with + 30 more

17,000+

developers

44,000+

CLI downloads

40+

coding agents

One workspace for your experiments

Drive the agents on your remote server, review what they did, and steer any of them, from your desktop or your phone.

Server

runs 2 agents

Mac mini

runs 1 agent

iPhone

controls

Web

controls

Laptop

runs 1 · controls

Vicoasecure relay

Drive the server your GPUs run on

Connect your GPU server, cluster nodes, or a cloud VM and drive the agents running on it through Vicoa. No SSH gymnastics, no VNC round-trip.

Your code and checkpoints stay on your own server
Reach every machine from your phone, laptop, or the web

Instead of SSH in every time you want to check a run

vicoa.ai/dashboard

Sessions

Add OAuth login flow synced

On it — I’ll wire up OAuth. Which providers?

Supabase + Google & GitHub

9:41

Add OAuth login flow

On it — I’ll wire up OAuth. Which providers?

Supabase + Google & GitHub

Type messages, @files, /skills…Supabase + Google & GitHub

Steer agents from your phone

Pick up the exact same session on your phone, approve a step, or redirect the agent, and it syncs everywhere instantly.

Approve the risky or expensive step from anywhere
Catch a broken run at 6pm, even when you are not at your desk

Instead of being chained to the terminal at your desk

Sessions 2 running 1 needs you

Add OAuth login flow

feat/auth

Fix cart total rounding

fix/cart

Document the REST API

docs/api

Migrate to Postgres 16

chore/db

Refactor billing module

refactor/billing

Agent team for parallel idea validation

Run several agents at once: one tries a new approach, another runs an experiment, another digs into the analysis.

Explore several ideas at once, each isolated on its own branch
See which agent is running, blocked, or done at a glance

Instead of one agent, one idea, one experiment at a time

Chat

Where should the token expiry live?

// tokens expire in 30 days

const exp = Date.now() + TTL

Add a comment…

// tokens expire in 30 days

const exp = Date.now() + TTL

src / auth /login.ts

auth/

jwt.ts

oauth.ts

18export async function login(user) {

19 const token = signJWT(user)

20 // tokens expire in 30 days

21 const exp = Date.now() + TTL

22 return { token, exp }

23}

Analyze your results

Open the file tree beside the conversation to inspect logs, metrics, checkpoints, and plots, then ask the agent to dig into what they show.

Browse outputs and artifacts without a separate transfer
Point the agent at a result and have it take the next step

Instead of scp-ing files back just to look at them

Review feat/auth

ChangesFilesTerminal

session.ts+6

oauth.ts+31

types.ts+2

src/auth/login.ts+18−3

17import { signJWT } from './jwt'

18export async function login(user) {

19− const token = signOld(user)

19+ const token = signJWT(user, opts)

20+ await audit(user.id, 'login')

21 return { token }

4 files changed+57−9 3 commits · reviewable on mobile

Review what the agent changed

Review every change as an inline git diff with word-level highlights, then flip to the files it touched or the terminal it ran, all in one view.

Per-file diffs and commit history
Verify the change instead of trusting a summary you can’t check

Instead of trusting a summary you can’t verify

Codex · fix/cart

Run the cart tests and fix the rounding bug.

Ran the suite — 2 files green, patching totals now.

editing cart/total.ts

zsh — fix/cart

➜ cart pnpm test

RUN v2.1.4 · 42 files

✓ auth/login.spec.ts (12)

✓ cart/total.spec.ts (8)

Test Files 2 passed · Duration 1.4s

db/migrate.spec.ts

➜ cart pnpm dev

A real terminal, on the server

Run commands, tail training logs, and watch a job in a live terminal on the same machine the agent uses.

Full shell access without leaving the session
Keep jobs, servers, and logs running while the agent works

Instead of juggling a second stack of SSH tabs

Cloud GPUs, lab servers, or your own hardware

Vicoa runs anywhere you can run a coding agent, and your codebase never leave that machine.

Lab & cluster

Your lab’s GPU server, or a node on the cluster.

Cloud VMs

Lambda, RunPod, Vast.ai, AWS, GCP, or Azure.

Your own hardware

A workstation or desktop at home.

From your server to your phone

Point Vicoa at the machine your experiments run on, in three steps

Install on your server

Install the Vicoa CLI on your lab server or cloud VM and sign in, and it auto-detects the coding agents already installed there.

Download for Mac

Prefer the terminal?

$npm i -g @vicoa/cli

Run experiments in parallel

Start agents on the server, each on its own git worktree, and see every session in one workspace.

feat/paymentsworking

fix/loginwaiting

mainidle

Steer from any device

Grab the mobile app to approve, redirect, and answer questions from anywhere.

Scan the code, or tap a store.

Vicoa Mobile

Your phone is the remote control

Approve changes, review diffs, and steer your agents from anywhere.

Reviewing a staged git diff in the Vicoa mobile app — Changes
Review changes agents made

Browsing the repository file tree in the Vicoa mobile app — Files
Browse and open any file

Reading an agent conversation in the Vicoa mobile app — Conversation
Read replies and reasoning in full

Vicoa mobile home listing coding sessions grouped by project — Agent Sessions
View your agents at a glance

Chatting with a coding agent in the Vicoa mobile app — Chat
Talk to any agent and approve changes

Lock-screen push notifications when Claude Code needs your input — Notifications
Get pinged when an agent needs you

Searching sessions, tasks, and automations in the Vicoa mobile app — Search
Jump to any session, task, automation

A task board with backlog, todo, and in-progress columns in the Vicoa mobile app — Tasks
Plan work for your agents

A list of scheduled automations in the Vicoa mobile app — Automations
Schedule agents to run themselves

Get The Mobile App

Install Vicoa on iPhone or Android

Open the same coding sessions from your phone with a real mobile app, not a terminal.

Scan to download the app

Already living in SSH + tmux?

Keep it. tmux keeps your session alive when you disconnect. Vicoa adds the workspace around it: run agents in parallel, review their work, and steer them from any device, at your desk or away from it.

What tmux does

Keeps your session alive.

Your job keeps running after you disconnect
Reattach from the same terminal

What Vicoa adds

Desk or away

A full workspace for your agents, on any device.

Run many agents in parallel, in one view
Review diffs, logs, and results as they land
Steer and approve from your desk or your phone
Get pinged when a run finishes or needs you

Who it’s for

ML & AI researchers

Training runs, evals, and demos on a remote GPU server, kicked off and steered without living in a terminal.

PhD & master’s students

Your experiments run on the lab cluster or a cloud VM; check on them and nudge them between classes, from any device.

Data scientists

Iterate on pipelines and models where the data and compute already are, and review the agent’s changes easily.

Labs & PIs

Equip the whole lab. When your students run more experiments and spend less time babysitting them, your lab produces more.

Loved by researchers

Featured on

Frequently Asked Questions

Answers for researchers running agents on remote machines

No. The coding agent runs on your own machine or server, so your code and checkpoints stay exactly where they already are. Vicoa relays your prompts and the agent’s output through a secure relay so you can drive that machine from another device. It relays the conversation, not your codebase.

40+ coding agents: Claude Code, Codex, OpenCode, Gemini, Cursor, Copilot, Kimi, and more built in, plus 30+ more supported via ACP, and 300+ models via OpenRouter. Bring your own subscription or API key. Vicoa does not resell model access, so you keep using the models you already pay for.

No. The agent runs on your remote server or VM, so it keeps working after you close your laptop. You check in and steer it from your laptop, phone, or tablet whenever you like.

It works best on a machine where you can keep a process alive: a dedicated lab server, a cloud VM, or an interactive cluster allocation.

Yes. You can share a running session so labmate can follow along and review the agent’s work, which is useful for feedback before anything lands.

Vicoa offers a free tier to get started, and you bring your own agent key, so there is no model cost added on top. For heavier use and advanced features, check the pricing page.

🚀 Join 17,000+ developers

Run a team of agents
on your server.

Set them up on your server, then steer, review, and approve from any device. Free to start with the agents you already use.

Easy setup

Works on any device

No credit card required

Run a team of coding agents
on your GPU server