Modules · Working with AI
In workYoo, AI doesn’t sit next to your work as a black box. It works as a team member in your normal flow: it gets tasks, asks in the chat, sets the status and waits for your approval. Everything is visible, everything traceable.
Optional · enabled per team · without activation no data goes to an AI
Claude · Stefan
Proposal in the task chat
Cause: The date is converted to UTC on export and ends up one day early.
Plan: Use the local date, add a test.
Risk: low · one file
Example. The project defines the buttons, only Stefan can press them.
How AI works in workYoo
Assign tasks, mention with @, reply in the chat. No separate tool, no copy and paste.
If something is unclear, the AI asks in the task chat instead of guessing.
With “My Claude”, only what you approve with a button gets implemented. Text in the chat is not enough.
AI log on the task, “working on it” notice, approvals in the security log.
“My Claude” sees exactly the tasks its person is allowed to see. No more.
Everything is a plugin, switched on per team. Without activation no data goes to an AI.
New · My Claude
Every person can connect their own Claude Code session, such as “Claude · Stefan”. It runs on their own computer, with their own repos and access. On the project you switch on “My Claude” and name the repo and branch, optionally for subprojects such as “Bugs” and “Features” too. Claude looks at new and changed tasks, asks in the task chat and makes a proposal with cause, plan and risk.
Flow on the task
1 · Question
“Which export shows the error, PDF or CSV?”
2 · Proposal
Cause, plan, files, risk and the project’s buttons
3 · Approval
“Approved: Implement + TEST (Stefan, 14:02)”
4 · Result
Message in the chat, status “internal review”
New · My Claude
Give your Claude meeting minutes or say “Write down the tasks for the marketing plan”. It creates tasks from that, with project, deadline and dates, “My date”, assignees, owners and priority. Before that it checks for duplicates, and if the project is unclear, it asks. Later it keeps the tasks up to date, moves a date or marks them done. This explicitly works in projects without code too, such as marketing or organisation.
Always labelled
Bot icon on the task
In list and card, in the detail view “Created by Claude · Stefan (AI)”
History and security log
Every creation and change can be traced.
Projects with AI marked
Bot icon in the project list and overview, faded when the AI only works in subprojects.
AI team members
With the “AI team members” plugin you create AI profiles that sit on the team like members. You assign tasks to them or mention them with @ in the task chat. They ask questions, write results into the description, set the status and assign themselves — but only with the abilities the admin enables per profile. The profiles run on the workYoo server with your team’s own Anthropic key.
AI assistant · beta
The AI assistant is a chat about tasks, projects, wiki and documents. The admin decides which data sources it may use. It has no access to time tracking or absences. The assistant is in beta and enabled on request.
How it works
Create “My Claude” in your profile and store the session key on your own computer.
Switch on “My Claude” on the project, name repo and branch, include subprojects.
Claude asks and proposes. You press the right button, or decline.
Instructions in the documentation. Claude works while your computer is running.
Data & security
AI assistant, AI team members and “My Claude” are three separate plugins. You can switch on one without the other. Without activation, no data goes to an AI.
All three use Claude by Anthropic in the USA. For the AI assistant and AI team members the workYoo server sends the data, for team members with your own key. For “My Claude” the workYoo server sends nothing to an AI: the session on the person’s computer works with their own Claude access.
With “My Claude”, a “just do it” in the chat or in a bug report does nothing. Every approval is in the security log with person, button and time. We store keys only as a checksum.
FAQ
No. All AI features are plugins and off by default. Your team decides whether and which to switch on. Without activation, no data goes to an AI.
AI team members run on the workYoo server with your team’s Anthropic key. “My Claude” is one person’s Claude Code session on their own computer, with their repos and access. They are two separate plugins: you can switch on one without the other, and one does not start the other’s bots.
It only implements when the person whose Claude made the proposal presses one of the buttons. Instructions in the chat or in a bug report do not count as approval. If the task changes afterwards, the proposal is reviewed again.
Yes. If you give it a note or an instruction in your session, it creates tasks and keeps them up to date later, also in projects without code. It only works in projects you can see yourself, only changes tasks it created and labels them with a bot icon. You can switch this off in your profile.
“My Claude” only sees tasks its person may open. AI team members work with task title, description, chat and project context. The AI assistant only uses the data sources the admin allows, never time tracking or absences.
To Anthropic in the USA. For AI team members, the workYoo server sends it with your team’s key. For “My Claude” the workYoo server sends nothing, the person’s session works with their own Claude access. Details are in the list of subprocessors and in the privacy policy.
Superadmins see every Claude on the team in the settings of the “My Claude” plugin. They can pause a Claude or block its key.
Colleagues whose own Claude watches the same project can take over an open proposal. This shows in the chat, the previous person is notified, and it is logged. An earlier approval no longer applies.
Start free and switch on the AI plugins when you’re ready.