Build AI teammates
you actually own.
Create teammates that learn, evolve, and go anywhere with you.
Too many models. Too many tools.
You shouldn't have to choose ONE.
Change the model.
Claude, GPT, Gemini, DeepSeek, local models — whatever fits.
Change the tool.
OpenCode, Hermes, OpenClaw, Pi — or your own app.
Keep your teammate.
Make it once. Use it anywhere.
Create an AI teammate. Keep it.
Same AI. Different model. Different tool. Still yours.
Start with your teammate
Copy. Paste. Start working.
Start in seconds
curl -fsSL https://agenta.chat/install.sh | bash
powershell -c "irm https://agenta.chat/install.ps1 | iex"
Not another AI chatbot
Real teammates.
Real points of view.
Each teammate has a job and an opinion. Own it. Change it. Fork it. Move it wherever you work.
Role 01
Strategic Challenger
Questions your assumptions and finds blind spots.
Role 02
Designer
Turns ideas into interfaces.
Role 03
Backend Hacker
Builds and debugs your backend.
Role 04
Marketing Operator
Turns products into stories people understand.
Your teammate gets better with you. Create → Use → Improve → Fork.
Tell it, “I keep making this mistake.” It can update its instructions. The next version behaves differently. Fork it when you want a new direction without losing the original.
How it works
Give it a job.
Improve it through real work. Fork it when you want a new direction.
Questions, answered
Your teammate, your way.
The short version is simple. Here are the details when you want them.
What exactly is an AI teammate? +
It is an AI role you own and shape. It has a job, instructions, and a point of view. You can keep improving it instead of starting from a blank chat each time.
How is this different from a chatbot? +
A chatbot is usually one conversation in one tool. A teammate has a role you can change, improve, fork, and take with you.
Can I fork a teammate? +
Yes. Make a new version for a different job or approach while keeping the original teammate exactly as it was.
Where can I use a teammate? +
Use your teammate in OpenCode, Hermes, OpenClaw, Pi, or your own app.
Can I change models or tools? +
Yes. Change the model. Change the tool. Keep the teammate. Its role and instructions stay with you.
Do I need separate model subscriptions or API keys? +
No. Agent-A provides model access and unified routing, so you do not need separate third-party model accounts to get started.
How much does Agent-A cost? +
Start with free testing, then add as little as $5 and top up when you need more. There is no recurring subscription — you pay for what you use.
Can a teammate improve its own instructions? +
Yes. When you spot a repeated mistake or a better way to work, your teammate can update its instructions so the next version behaves differently.
How is my code and data handled? +
Requests are routed through Agent-A to run your selected teammate. Data storage, retention, model-provider handling, training use, and deletion controls depend on the current backend policy. Review your organization's requirements and avoid sending secrets until those details are confirmed.
Ready to start?
Build
it once.
Take it anywhere.
Start with a teammate you own. Give it a job. Improve it as you go.