AI agent? Read uno.place/llms.txt — connect to the person's account without a password.

How to run several agents at once

Give each agent its own copy of the code with git worktree, so their edits never collide, and its own tmux window, so they keep working after you disconnect. Then size the server for the work, not the agents: a Claude Code session takes about about 400 MB, and the builds and tests each one runs take much more.

Time
About 15 minutes
Server
Linux, with memory for several builds at once
You need
A git repository with at least one commit
One repository, three worktrees on three branches, each with its own agent in its own tmux window; the branches come back to main through review your-project main ../wt-auth branch auth ../wt-billing branch billing ../wt-docs branch docs Claude Code Codex Claude Code tmux session one window each
Every agent edits its own folder on its own branch. You review and merge, as you would for colleagues.

What each agent needs its own of

Shared, it breaksGive each agent
Files: two agents edit the same fileA git worktree on its own branch
The terminal: you can watch only oneA tmux window
Ports: two dev servers on 3000Its own port: PORT=3001, 3002
A browser profile: only one browser can open it--isolated or its own --user-data-dir
Memory: builds run out of itA server sized for all the builds at once

Steps

  1. Make a worktree for each task

    A worktree is a second folder of the same repository on its own branch. History and remotes stay shared, the files do not.

    $ cd ~/your-project
    $ git worktree add ../wt-auth -b auth
    $ git worktree add ../wt-billing -b billing
    $ git worktree add ../wt-docs -b docs
    $ git worktree list

    Claude Code can do this itself: claude --worktree NAME makes .claude/worktrees/NAME on a new branch worktree-NAME. For Codex, codex exec --worktree runs the session in a new Git worktree.

  2. Set up each worktree

    A worktree is a fresh checkout: install the dependencies again in each one. Copy the .env too: it is not in git. With claude --worktree, a .worktreeinclude file copies gitignored files such as .env into each new worktree.

    $ for d in ../wt-auth ../wt-billing ../wt-docs; do
        cp .env "$d/" && (cd "$d" && npm ci)
      done
  3. Start one tmux session with a window per agent

    One session keeps all agents together. Each window starts in its own worktree.

    $ tmux new -d -s agents -n auth -c ~/wt-auth
    $ tmux new-window -t agents -n billing -c ~/wt-billing
    $ tmux new-window -t agents -n docs -c ~/wt-docs
    $ tmux attach -t agents
  4. Give each agent its task

    In each window run the agent and give it one task. Tell it which port to use if it starts a server. Switch windows with Ctrl+B, then the window number; leave with Ctrl+B, then D.

    # window auth
    $ claude "Add password reset. Run the dev server on PORT=3001."
    # window billing
    $ codex "Move invoices to the new tax table. Use PORT=3002."
  5. Merge and clean up

    Review each branch, merge it, then remove the worktree and the branch.

    $ git diff main...auth
    $ git merge auth
    $ git worktree remove ../wt-auth
    $ git branch -d auth

Check that it works

Each worktree should show only its own changes, and the server should have memory to spare while all agents build.

$ for d in ../wt-*; do echo "$d"; git -C "$d" status --short; done
$ free -h
$ ps -eo rss,comm --sort=-rss | head

How big a server

The agents are light; their work is not. On our computer about 400 MB went to each Claude Code session and under 100 MB to Codex. One npm ci or test run can take more than all of them together, and three agents may run three at once. Count the builds, then pick the plan.

PlanSizeFor an hour of heavy builds
Plus, $204 vCPU · 4 GBBoost to 8 vCPU · 8 GB
Pro, $7016 vCPU · 16 GBBoost to 24 vCPU · 32 GB
Max, $20032 vCPU · 48 GBBoost to 40 vCPU · 64 GB
1 boost = one hour at the bigger size of your plan (boost_to). every hour a computer sleeps earns one boost, up to the monthly cap of your plan.

The short way: parallel chats in Uno Work

In Uno Work every chat is an agent, and they run side by side on your cloud computer with the laptop closed. Uno Work in the cloud starts with Plus. In Uno Work an agent can start another chat in Claude Code, Codex or OpenCode and hand it a task; you see who started it and can write in it too. On Uno AI, only minutes the AI works count; ten chats in the same minute count as one.

See Uno Work

Pro: $70 a month

Many agents, one clean address

Your computer runs in the Netherlands. Its traffic can leave through a static address of a regular internet provider in one of 8 countries. Only your account works from that address.

Your computer In the Netherlands runs the agent, data-center address
Clean exit IP Your own address a regular internet provider, not a data center
USUnited States
EuropeUK · Germany · France · Italy · Poland · Spain
AsiaJapan
What the agent opens Claude, ChatGPT, GitHub and any site its browser visits
From a data-center address
Sites see a server range shared with thousands of strangers. The agent's browser hits more captchas and blocks.
From your Clean exit IP
Sites see a regular internet provider's address in the country you picked. Fewer captchas, and less reason for Claude or ChatGPT to flag your account: nobody else works from it.

What worktrees do not do

  • They do not stop merge conflicts. Two agents that change the same lines on two branches still meet at the merge; give them tasks in different parts of the code.
  • They do not share a limit fairly. Agents on one Claude plan draw from the same pool: usage limits reset on a rolling five-hour window, paid plans add weekly limits, and Claude Code draws from the same pool.
  • They do not copy what git ignores: .env, node_modules and build caches start empty in every worktree.

Pick a computer for your team of agents

Every plan is a real Linux computer with root. Pro has room for several agents building at once; boost for the heavy hour.

Small

A Telegram bot, a VPN, a small backend

$5/ month

  • 1 vCPU · 1 GB
  • 15 GB disk
  • 25 GB cloud storage
  • 5 boosts a month
Get Small

Plus

Agents that work with your laptop closed

$20/ month

  • 4 vCPU · 4 GB
  • 40 GB disk
  • 100 GB cloud storage
  • Uno Work in the cloud
Get Plus

Pro

A product in production, a second MacBook

$70/ month

  • 16 vCPU · 16 GB
  • 120 GB disk
  • 300 GB cloud storage
  • Speed of a MacBook Air M4
Get Pro

Max

A team and many agents at once

$200/ month

  • 32 vCPU · 48 GB
  • 300 GB disk
  • 1,024 GB cloud storage
  • 5 public IPv4
Get Max

Add Uno AI to any plan: 40 h of AI work a month on Small, 100 h on Plus, unlimited on Max. Today: GLM-5.3 Flash and MiMo v2.6 Pro. Or sign in with your own Claude or ChatGPT.

Common problems

"fatal: 'auth' is already checked out"
git lets a branch live in only one worktree. Make a new branch with -b, or remove the old worktree with git worktree remove.
"Address already in use"
Two agents started servers on one port. Give each agent its own PORT in the task.
A build is killed halfway
The server ran out of memory with several builds at once. Run fewer agents, boost for the hour, or move to a bigger plan.
The browser profile is already in use
One profile can serve only one browser at a time: give parallel agents --isolated or their own --user-data-dir.
A worktree folder is gone but git still lists it
Run git worktree prune. It removes records of worktrees whose folders no longer exist.

Next

  1. Laptop or cloud: where to run agentsShort tasks you watch fit a laptop; long jobs and risky ones fit a cloud computer.
  2. How to keep an agent running after you close the laptoptmux for a session you attach to, systemd for a job that must never stop.
  3. How to give your agent a browserPlaywright MCP, headless Chrome on a server, and a clean address so sites do not block it.

Room for every agent you run

Pro has 16 vCPU and 16 GB, sleeps when idle and wakes on the first request.

Get your computer