Team AI ConsoleESDownload free

Desktop app · uses your Claude Code or Codex

One AI agent for every role, always in your orbit.

Team AI Console is an app you install on your own computer: the work is done by the Claude Code or Codex you already have, with your subscription. The app adds what they are missing: roles with their responsibilities, turns and documents. You define your own roles, teams and workflow, and you lead them on your own or share them out across your team.

Download it free

No card, just your email.

No API key. No server of ours in between.

Read the documentation

On your own computer

It is not an online service: it is an app that uses your own AI agent.

You install Team AI Console on your computer and the work is done by the Claude Code or Codex you already have, with your subscription. You pay no tokens to any service of ours.

  • Your computer

    The app, your projects and their documents live on your disk. macOS, Windows and Linux.

  • Your subscription

    It uses the Claude Code or Codex you already pay for. No API key, and no tokens we charge you for.

  • Your repository

    The team’s work travels through your own git: GitHub, GitLab, Bitbucket… No central server.

What it is

  • An app you install on your own computer.
  • It uses the Claude Code or Codex you already have.
  • Method, roles and turns on top of the agent you use.
  • Documents and code, inside your own project.

What it is not

  • A web service you upload your code to.
  • An AI provider that charges you for tokens.
  • An AI model of its own.
  • A platform that stores your project.

  • Your computer

    The app, your project, its documents and the Claude Code or Codex you have installed.

  • Your AI provider

    Anthropic or OpenAI, with your account and your subscription, just as if you used the agent directly.

  • Our server

    Only the license: key or ID, a non-reversible fingerprint of the machine, a readable name, the app and system versions, and your email, only when you request the license.Never your code or your conversations.

AI doesn’t lack talent. It lacks method.

AI agents already know how to code. But used raw on a real project, the same problems show up every time.

  1. It jumps straight to code

    You ask for a feature and it writes code without asking who it’s for, what exactly it should do, or how you’ll know it works.

  2. It makes up what it doesn’t know

    It fills the gaps with assumptions nobody validated: requirements, data, product decisions.

  3. It leaves no trail

    Everything lives in a chat that gets lost. A week later, nobody knows why something was done or who approved it.

Who it’s for

How it works

Four ideas human teams already use. Now applied to AI.

  1. 01

    Roles

    Each team has its own: Product Manager, Business Analyst, Designer, QA… Each one has its own document: who it is, what it can do and what it can’t. And you can create your own roles.

  2. 02

    Turns

    Only one role works at a time. When it’s done it proposes who’s next, “Switch to: Business Analyst”, and you accept the handoff.

  3. 03

    Flow

    Turns follow the path each team defines, like a sprint, or the one you design yourself. The app draws it and marks which step the team is on.

  4. 04

    Documents

    Every step leaves a Markdown deliverable in your project. It’s the project’s memory and the next role’s input.

An example: one team’s flow, 15 steps in 5 phases

Steps and phases come from each team’s flow · you can define your own
  1. PHASE 1 · 1 STEPKickoffThe Product Manager sets the sprint goal and context.
  2. PHASE 2 · 4 STEPSDefinitionThe Product Owner defines the feature; the Business Analyst writes requirements and acceptance criteria.
  3. PHASE 3 · 5 STEPS · NOWPlanningSprint plan, design and mockups, test cases and technical approval.
  4. PHASE 4 · 3 STEPSBuildThe Developer implements, QA runs the tests and the Product Manager validates the delivery.
  5. PHASE 5 · 2 STEPSCloseThe Documentation Lead documents; the Scrum Master closes the sprint.

Nothing moves forward without a person: someone with the same profile, a BA or a QA, reviews what their agent does and approves it or asks for changes. Or one person leads the whole team.

You approve each deliverable and each gate, not every action: within a step, the role implements, tests and fixes on its own.

Working with others

Several people, one project. No central server.

Each person installs Team AI Console on their own computer and takes one or more roles. Everyone works on the same project, shared through git. When someone finishes their turn, the app notifies the person holding the next role.

On your own, or shared out across several people

  • One computerOn your own

    You take every role on the team: you talk to each one, read its deliverable, approve it and pass the turn to the next. You need nobody else.

  • One computer per personWith your team

    Each person installs the app and claims the roles of their profile. The app passes the turn, notifies whoever is next, and everything syncs through your git repository.

  • AnaHer laptop · macOS
    Product ManagerTurn handed off
  • SandalioHis PC · Windows
    DeveloperOn their turn
  • MartaHer PC · Linux
    Business AnalystWaiting for their turn
SandglassSame project · shared git repository

When it’s not your turn, you read everything but don’t write.

  1. Everyone picks their roles

    From their own app, each person claims the roles they own. One person can hold several.

  2. One project for everyone

    Documents, decisions and code live in the same git repository (GitHub, GitLab, Bitbucket…), with no central server. Every turn is saved as a commit under the name of whoever ran it: a history you can hand to your client.

  3. The turn passes from person to person

    When done, the app notifies whoever is next and hands over the context.

Real case · Sandglass

“show a timer on screen, with an hourglass animation, that’s all”

One sentence. 13 documents. A working app.

From that single request, one of the teams produced a complete, traceable sprint, with every decision explained and approved.

  1. 01Product brief
  2. 02Requirements
  3. 03Scope review
  4. 04Sprint plan
  5. 05Interface spec
  6. 06Test plan
  7. 07Technical design
  8. 08Code change
  9. 09Code review
  10. 10Test report
  11. 11Delivery acceptance
  12. 12Retrospective
  13. 13Closure

Features

Always know who’s working, which step they’re on and what they delivered.

  1. Team of roles

    Each task is done by the right specialist, with its own rules and limits. One click activates a role.

  2. Suggested handoff

    When a role finishes, it proposes the next one. The process moves on its own, but never without your OK.

  3. Flow map

    The flow’s phases and steps, with the current step marked and each gate’s paths: “All good” moves on, “There is a problem” goes back. You always know where the project is and what’s left.

  4. Team documents

    Built-in Markdown viewer: read each role’s deliverable without leaving the app. It updates itself and opens the latest.

  5. Usage

    See how your team works: tokens, turns, rework and context re-read, by role, step, chat, model and day, plus the API equivalent with Claude Code. Save a summary for the retrospective. An indicator, not a bill. Not included in the free license.

  6. Delegate

    A role of the team decides on your behalf, with your criteria, and the sprint moves on by itself. Three autonomy modes: Supervised (you decide), Semi-autonomous (the delegated role decides most steps and the flow’s reserved steps wait for you) and Autonomous (it decides them all). It stops for questions and permissions, it always asks you before pushing to git, deploying, deleting data or paying, and every decision is recorded for you to review.

  7. Your backlog from GitHub or GitLab

    The project backlog can come from your repository’s open issues: one button brings them in, one item each, with their title and description, with no duplicates. The team starts the sprint with the ones you pick and, when done, offers to close each issue with a comment that links to the sprint’s closing.

  8. Review and annotate documents

    Select a fragment of a role’s document and comment on it, suggest a replacement or mark it for deletion: “Request changes” sends the role a structured request, with the section and fragment of each annotation, and then you see which ones it addressed. The Changes, History and Compare tabs show which files each turn touched and the document’s versions, with their differences.

Templates

Ready-made teams for every situation. Or your own.

The app ships with several teams, each with its own roles, workflow and documents. Pick the one that fits what you need, build your own in the app, or import one.

  • Asteroid1 role

    Ceres

    You have an idea and want to pin it down.

    One role that turns a vague idea into a Vision Document.

  • Nebula5 roles

    Orion

    You want to define the product before coding it.

    Requirements, interface design, test cases and technical design, ready to implement.

  • Supernova9 roles

    Vela

    You want to build and test the product.

    Planning, code, testing and security, all from the app.

  • Galaxy10 roles

    Andromeda

    You want to take it all the way to production.

    Everything above, plus deployment, servers and operations.

Your own team

Import a team created outside the app. Team AI Console reads the structure of its folder and lets you use it like any of the teams above. The folder holds:

  • A document for each role
  • The workflow, described in a document
  • A configuration file

Create your own team

In Teams and Roles, on the app’s Home, you build your own: an agent interviews you about how you work and writes each role’s guide with you; you define the team’s roles, its flow of phases and steps and its documents, try it in a test project and export it as a .tacteam file.

When the team improves, the projects that use it get a new-version notice when they open. Creating roles and teams isn’t included in the free license.

Frequently asked questions

License

Free. No card.

A free license, with some features not included, such as creating roles and teams. One license per machine; you can move it to another. Your AI agent uses its own subscription.

Is it an online service? Do I pay for tokens?

No. Team AI Console is an app you install on your own computer, and it works with the Claude Code or Codex you already have, with your subscription. There is no AI service of ours in between and no tokens we charge you for: only the app license, and the free one asks for no card.

Do I need an API key?

No. The app uses the agent installed on your computer with your current subscription.

Does my code leave my computer?

Team AI Console doesn’t send your code to any server of its own. The AI agent you choose talks to its provider just as it would if you used it directly. Collaboration travels through your own git repository.

Does it work for teams?

Yes. Each person claims the roles that match their profile (a BA, a QA…), reviews what their agent does and approves it or asks for changes. The app manages turns between people and syncs over git, through your own repository (GitHub, GitLab, Bitbucket…).

Do I need one person for each role?

No. You can share the roles among several people, each with the profile of the role they review, or lead the whole AI team on your own. One person can hold several roles.

Do I need to know how to code?

Not to lead the team: you talk to the roles, read their documents and approve. You need Claude Code or Codex installed and, to collaborate, git. Knowing how to code helps when reviewing the developer role’s work.

Which AI agents does it support?

Claude Code and Codex CLI. Each role can work with either one: if you have both, you choose each role’s agent. Each role can also use its own model, for example a more powerful one for design and code and a faster one for the rest. Choosing the agent and model per role isn’t included in the free license.

Why not use Claude Code or Codex directly?

Team AI Console uses them: it doesn’t replace your agent, it organizes its work. It adds a method (a team’s flow of phases, steps and gates), responsibilities (each role with its rules and limits), verifiable deliverables (each step leaves a document in your project) and coordination (turns, handoffs and, if you work with others, roles shared among people over git).

Can I see how much each role costs?

Yes. The Usage view shows tokens, turns, rework and context re-read by role, by flow step, by chat, by model and by day and, with Claude Code, the API equivalent at list price. With a subscription plan it isn’t real money: it’s an indicator, not a bill. It only counts your own work in the app and isn’t shared through git. Not included in the free license.

What data does the app send?

It only talks to the license server, with the minimum data needed to validate the license: the license key or ID, a non-reversible fingerprint of the machine, a readable machine name (user@computer), the app and system versions, and your email only when you request the license. It never receives your code or your conversations.

Is there just one AI team?

No. The app ships with several teams, each with its own roles, flow and documents: from a single-role team that pins down an idea to one that takes the product to production. The roles and the sprint on this page are examples. You can also create your own teams in the app, or import one.

Is it only for building software?

No. Ceres and Orion produce documents: the vision, the requirements, the design and the test cases, so another team can build the product.

Can I change the roles or the flow?

Yes. The app doesn’t fix the roles, the teams or the method: in Teams and Roles you create your own roles and teams, with their flow of phases and steps and their documents, with an agent that interviews you about how you work, without coding. You can start from a copy of a team that ships with the app. Each team is kept in text documents and a config file inside the project, so you can also edit them by hand.

Do I have to approve every step?

No. You approve deliverables and gates, not each action: within a step, the role works on its own. If you don’t want to be there for every confirmation, with Delegate a role of the team decides on your behalf, with your criteria; it records every decision and still asks you before pushing to git, deploying, deleting data or paying. There are three autonomy modes: Supervised (you decide everything the flow reserves for you), Semi-autonomous (the delegated role decides most steps and the flow’s reserved steps still wait for you) and Autonomous (it decides every step). Which steps are reserved is set by the team in its flow. Permissions set what the agent asks before acting, with each file change shown line by line: always ask, accept file edits or automatic.

Can I bring in my GitHub or GitLab issues?

Yes. In the project’s Backlog, “Import issues…” brings in the open issues of the repository’s remote (github.com or gitlab.com) and creates one pending item per issue, with its title, its description and a link to the issue; the type comes from the labels (bug, feature or change). You can import again without duplicating anything; public repositories need no credential and private ones use the one git already has saved. When an imported item is done, the app offers to close the issue with a comment that links to the sprint’s closing, and only closes it if you confirm.

Does the team remember what was done?

Yes. Each chat is its own agent session and is resumed when you reopen the app. Between chats and team versions, the shared context is the living documents in your project: a new chat starts from them, not from the previous conversation.

Does it work for a large company?

For definition teams and small projects, yes. It doesn’t replace a large company’s engineering process. Work travels through your git repository (GitHub, GitLab, Bitbucket…); it doesn’t integrate with Jira or Slack.

Which languages is it in?

Spanish and English. The app works in the team’s language.

What happens if my license expires?

The app switches to read-only: you can open your projects and read every document. Nothing is lost.

Does it work offline?

The app works without validating the license for a grace period. The AI agent does need a connection to its provider to work.

Every role takes its turn.Every decision is yours.

It’s free. No card, just your email.

Requires Claude Code or Codex CLI with an active session, and git to collaborate.

Installation and first steps, in the documentation

First time on a Mac? How to open the app

Download the app

Available for macOS, Windows and Linux. If the download links don’t appear, write to us: Talk to us.

macOS

Opening the app the first time

The app isn’t notarized by Apple yet, so macOS warns you the first time you open it. You only need to allow this app; you don’t need to change any other security setting.

  1. Step 1

    Open Team AI Console from Applications. macOS says it couldn’t verify the app: click Done, not Move to Trash.

  2. Step 2

    Open System Settings › Privacy & Security.

  3. Step 3

    Scroll down to the Security section and, next to the message about Team AI Console, click Open Anyway.

  4. Step 4

    Confirm with Touch ID or your password. If macOS asks once more, click Open Anyway. From then on, the app opens normally.

On macOS 14 Sonoma or earlier it’s quicker: right-click the app in Applications, choose Open and confirm with Open.

Repeat these steps after installing each new version.

Illustrations of macOS 15 Sequoia and later. The exact wording depends on your macOS version and language.

Talk to us

Want a demo or a licence for your team? Tell us about your agency and we’ll reply by email.

Fields marked * are required.

This form needs JavaScript. Turn it on in your browser to send us a message.