New in 1.8 Swarm, Chat and the Dashboard

One room for every agent your team runs.

Your developers already work with Claude Code, Codex and Gemini. SpaleStudio puts all of them side by side in one Windows workspace, so the whole team sees what each agent is doing, what it changed and what it is waiting for.

Windows 10 and 11 · Agents keep the accounts you already pay for

Real screenshots · SpaleStudio 1.8
SpaleStudio with four agents in the Release 4.2 workspace: Thea and Reed, both Claude Code, report what they changed; Luna, also Claude Code, asks whether to make an edit to schema.mjs; Leo, in Codex, has written the refund tests. A notification says Luna needs permission to continue.
  1. Every agent by name. Thea, Reed and Luna run Claude Code, Leo runs Codex, all in one workspace.
  2. Who needs you. SpaleStudio says which agent is waiting for an answer, even from another workspace.
  3. Nothing happens behind your back. Luna asks before it edits schema.mjs, and waits.

It launches the agents your developers already use

Every tile is a real terminal in your project folder, so whatever else your team runs in a terminal runs there too.

  • Claude Code
  • Codex
  • Gemini CLI
  • Cursor
  • Copilot
  • OpenCode
  • Kimi
  • Grok

The story

The agents got good. The desk didn't.

Not long ago an AI agent was a chat box. Today it reads the repository, writes the migration, runs the tests and asks before it deletes anything. Most developers keep three or four going at once, in a pile of terminal windows.

Nobody can see which one is stuck on a question or what was decided yesterday, and when a window closes, its context goes with it. So we kept the agents exactly as they are and built the room around them.

Before Nine terminal windows
After One SpaleStudio window
The same work in SpaleStudio: four agents in one window, and a notification that one of them needs permission.

Every agent in view The one that needs you, flagged Every change in one place

  • 8 agent CLIs it launches, from Claude Code to Grok
  • 12 agents can share one goal in a Swarm
  • 1 installer for every Spale program
  • 0% of your model spend goes to Spale

A day in SpaleStudio

One day in that room.

A release day on a demo billing service. The pictures are real screenshots of SpaleStudio 1.8 with real agents doing the work; only the project is made up.

  1. One goal, one team.

    The lead starts a Swarm with one goal: work refunds out from the ledger for release 4.2. A coordinator, a builder and a reviewer take their places, and every one of them works to the same rules: a failing test first, and a review of its own diff.

    The Swarm Refunds on the ledger: Coordinator 1 linked to Builder 1 and Reviewer 1, with the toolbar for roster, plan, chat, activity, memory and Git.
  2. The plan lands on the board.

    The coordinator reads the code and writes the plan as tasks on the project's board: derive refunds from ledger_entries, backfill the historic ones, test every refund path, and write the changelog entry for the support team.

    The billing-service board with four tasks to do, written by the Swarm's coordinator.
  3. A question waits for a person.

    The Swarm runs safe, so the builder asks before it runs a command, and SpaleStudio puts the question in front of you. Messages from other agents wait until you have answered, so nothing picks an answer for you.

    Builder 1: permission request. Choose a permission: review this option in the terminal, or deny this operation.
  4. Every change, before it ships.

    Four agents worked in the same folder. The Git panel lists every file they touched, shows the diff, and lets someone review it before anything is committed.

    SpaleStudio's Git panel on branch main: five changes, among them schema.mjs, refunds.mjs, the new 0042_ledger_entries.sql and refunds.test.mjs, with the diff of schema.mjs.
  5. Where everything stands.

    The Dashboard shows every open workspace, each agent in it and what it is doing or waiting for: the Swarm's three agents, and Thea, Reed, Luna and Leo in Release 4.2.

    The Dashboard: the Swarm Refunds on the ledger with Coordinator 1, Builder 1 and Reviewer 1, and the Space Release 4.2 with Thea, Reed, Luna and Leo, each with its last action.

Features

Everything the room needs.

The rest of SpaleStudio 1.8, beside the terminals your agents already run in.

Swarm

A team of agents on one goal, from a builder and a reviewer up to twelve, working to rules you switch on once.

Setting up a Swarm: the rules step, with Test-Driven and Code Review switched on.

Dashboard

Every workspace you have open, every agent in it, and what each one is doing or asking right now.

The Dashboard with SpaleSwarm, SpaleSpace, SpaleBoard and SpaleMemory side by side.

Chat

Talk to Claude, Codex or another installed CLI with files, pictures and sources, and switch CLI without losing the thread.

Routines

Work that starts on its own, at times you set like an alarm.

Focus

See where each agent is working, and follow the one writing, letter by letter.

Plugins and Addons

MCP servers, skills and hooks for your CLIs in one step, with keys kept out of source files.

Browser

A browser inside SpaleStudio. Links from a terminal open in it, beside the code.

Memory

A shared vault of linked notes your agents search and write to, with a record of who changed what.

The programs

Three programs, one account.

SpaleHub is the one download. Sign in once on a computer and it installs, licenses and updates the rest.

SpaleStudio with four agents side by side.

SpaleStudio

Where the agents work

Agents side by side in real terminals, with Swarm, Chat, Board, Memory, Routines, a browser and Git around them.

SpaleHub

Where the licences live

Signs each computer in, installs the other programs and keeps them updated with signed releases, add-ons and release notes.

SpaleVoice

Talk instead of type

Press a key or a mouse button, speak, and the words are typed into whatever field has focus. Use your OpenAI key, or a Whisper model on the computer so your voice never leaves it.

For companies

Built for your company's computers.

What a company should know before it puts Spale on its developers' machines, and the page it manages them from.

  • Code stays in your folders

    SpaleStudio opens project folders on each developer's computer and runs real terminals in them. Spale doesn't upload your projects.

  • Agents bill through your accounts

    Each agent signs in with the provider your company already pays. Spale doesn’t resell model usage or take a cut.

  • A licence is a computer

    Your account page shows which computers use each licence. When someone leaves, remove theirs.

  • A new computer needs a code

    Adding a computer takes a six-digit code sent to the account's email, so a leaked password can't take a seat.

  • Every update is signed

    SpaleHub installs a release only when its signature checks out, for every Spale program.

  • Agents ask first

    Each workspace starts safe or with Skip Permissions; you choose. While an agent asks, nothing answers for you.

Your account page on this site, with an example company: who holds each licence, and on which computers.

How a company starts

  1. Create an account One account holds your company's licences.
  2. Install SpaleHub on each computer Windows 10 or 11, 64-bit. It is the only installer.
  3. Confirm each computer Sign in; a six-digit code goes to the account's email.
  4. Install SpaleStudio and SpaleVoice From inside SpaleHub, which keeps them up to date.

The Spale plan

One plan. No tiers.

Every Spale program on one account. The price isn't set yet: until it is, licences are issued by hand, and you'll hear the price before anything is charged.

  • SpaleStudio, SpaleVoice and what Spale ships next
  • One seat per computer
  • Updates, add-ons and skills through SpaleHub
  • Your own agent accounts, with no cut taken

Questions

Questions companies ask.

Does our code leave our computers?

Spale doesn’t upload your projects. The agents talk to their own providers, just as they do outside SpaleStudio, under the terms your company has with those providers.

What do we pay for model usage?

Nothing to Spale. Each agent bills through its own provider account. Spale doesn’t resell usage and takes no cut of it.

Which agents can we use?

SpaleStudio launches the command-line agents installed on the computer: Claude Code, Codex, Gemini, Cursor, Copilot, OpenCode, Kimi and Grok.

Can agents act without asking?

Only if you let them. Each Space or Swarm starts either safe, where agents ask before they act, or with Skip Permissions. You choose every time you start one.

What counts as a seat?

A computer. SpaleHub licenses the Spale programs on the computer it runs on and renews the licence while it is open. SpaleStudio pauses if SpaleHub closes and carries on when it is back.

Which systems are supported?

Windows 10 and 11, 64-bit. If a Mac or Linux version is made, the download page will say so first.

Download

Give your team's agents a room.

Install SpaleHub, sign in, and open your first project with every agent in view.

Windows

Windows 10 or 11, 64-bit. SpaleHub is the one installer: it signs the computer in, then installs and updates SpaleStudio and SpaleVoice.

Download SpaleHub

Every release is signed. Release notes

Your own agents

Each agent signs in with the provider your company already pays, the way it does outside SpaleStudio. Spale takes no cut of that usage.

  • Claude Code
  • Codex
  • Gemini CLI
  • Cursor
  • Copilot
  • OpenCode

Mac and Linux

Not yet

Spale is a Windows program today. If a Mac or Linux version is made, this page will say so first.

Create an account