JustTalk AI lets you build voice-driven AI agents that your students practice with inside the JustTalk mobile app between live lessons. Agents have custom personalities, knowledge bases, and evaluation criteria.
Click Create Your First Agent (or + New Agent) to open the builder.
The agent builder
The builder has five tabs:
| Tab | What it controls |
|---|
| Agent | Name, category, description, system prompt, first message, default personality toggle. |
| Knowledge Base | Files and references the agent can draw on (lesson plans, reading texts, vocabulary lists). |
| Analysis | Evaluation criteria the agent uses to score the student’s performance. |
| Students | Which students on your roster have access to this agent. |
| Test | Try the agent yourself before you assign it. |
Default system prompt
The default prompt gives you a complete English teacher persona (“Professor Anya Sharma”) that you can customize or replace entirely. Type {{ in the prompt field to insert variables (student name, level, goals).
Agent settings panel
On the right side of the builder:
- Agent Name — shown to students.
- Category — e.g. Conversation Practice.
- Description — short blurb shown on the agent card in the mobile app.
- How to use — optional, longer guidance shown to students on the agent page.
Workflows
The second tab at the top of the JustTalk AI page (Workflows) lets you build multi-step agent flows — for example, a placement test that branches into different practice scenarios based on the student’s level.
Best-practice agents
A great first agent is a scenario role-play (restaurant, job interview, customer support call). Pin a clear objective in the first message and a single evaluation criterion in the Analysis tab — short, focused practice beats sprawling chats.
- Scenario role-plays — restaurant, hotel check-in, doctor’s office, job interview.
- Pronunciation drills — read-and-repeat with feedback against a target accent.
- Vocabulary stretchers — agent prompts that force the student to use a specific Focus Set word.
- Homework review — student summarizes a reading; agent grades and asks follow-ups.
