Talk to one.
Get a whole company.
Girru gives you a team of AI employees — each with a role, skills and permissions — that remember your work and act only with your approval.
Remembers
Every reply shows what it remembered. You approve what it learns.
A real team
One agent hands each task to the right specialist. A shared task board shows who does what.
Yours
Runs on your computer with the Claude or ChatGPT subscription you already have, or local models.
How it works
Three steps, then it runs.
You set the company up once. After that you talk to one agent, and only step in when something needs your OK.
- 01
Build your company
A short wizard sets up departments and the specialists you need.
- 02
Talk to one agent
It hands each task to the right specialist and keeps you posted.
- 03
Approve what matters
Each agent has its own permissions. Set money, deletes and messages to ask you first.
Memory · coming in 1.4
It remembers. You stay in charge.
Memory you can see, approve, correct and forget — stored on your computer.
What it remembered
Every reply shows the memories it used, so you know why it answered that way.
Speaks your language
Understands Arabic, including the Iraqi dialect — not only English.
You approve what it learns
It proposes, you decide. Conflicts with older memories are flagged.
One memory for all your tools
The same memory reaches your coding tools through project files and MCP.
Agents as employees
Each one has a job — and limits.
Give every agent a role, skills and its own permissions. Handing work to another agent never gives it more power.
- Personas & specialties
- A name, a role and the skills it needs.
- Task board
- Company, department and shared boards — board, table, calendar and workload views.
- Scheduled tasks
- Recurring work runs on time, with the same permissions.
- Girru Store
- Add apps, skills and specialties — checked before they are set up.
Your subscription, your data
No new bill. No cloud copy.
Girru uses the AI subscription you already have, or local models. Your company, memory and files stay on your computer.
- Claude or ChatGPT
- Uses your subscription through Claude Code or Codex on this computer.
- Local models
- Ollama or LM Studio — run fully offline if you prefer.
- On your computer
- No tracking, nothing stored on our servers. Prompts go only to the model you choose.
Preview and coming next
What we are building now.
3D worlds with GenVR
Agents that build 3D worlds. Early preview, still changing.
Trading, paper mode
Practise trading with paper money. No real money is connected.
SEO app
An SEO app inside your company, next to your SEO specialist.
Phone
Talk to your company from your phone.
Releases · building since May 2025
Next: Girru 1.4
- Every reply shows the memories it used.
- It proposes, you decide; conflicts with older memories are flagged.
- Understands Arabic, including the Iraqi dialect.
- The same memory reaches your coding tools through project files and MCP.
Download
Install in two minutes.
- Open the .dmg and drag Dpro to Applications. The app is not signed yet: the first time, right-click it and choose Open.
- Start Dpro and create your owner account on this computer.
- Open Agents → Build your company to hire your first team.
- Run the installer.
- Start Dpro and create your owner account on this computer.
- Open Agents → Build your company to hire your first team.
- Make the AppImage executable and run it.
- Start Dpro and create your owner account on this computer.
- Open Agents → Build your company to hire your first team.
FAQ
Questions.
Is it free?
Yes. The app is free and open source (MIT). You only pay for your own AI subscription, if you use one.
Where is my data?
On your computer. Your company, memory and files are not uploaded to us. Your prompts go only to the model you choose.
Which models does it use?
Claude through Claude Code, ChatGPT through Codex, Gemini, local models through Ollama or LM Studio, or your own API key.
Do I need an API key?
No. If Claude Code or Codex is signed in on your computer, Girru uses that subscription. Local models need no key either.
Is it safe?
Every agent has its own permissions (never, ask me or allow), and handing work to another agent never gives it more power.