Skip to main content

​
What is Agent Mode?

Agent Mode is your AI-powered assistant inside OptiDev. Instead of writing code yourself, you describe what you want in plain English, and the agent builds it for you. Think of it as having a developer on your team who’s always available. You can ask it to:
  • Build new features (“Add a contact form with email validation”)
  • Fix problems (“The button isn’t working, can you fix it?”)
  • Explain how things work (“How does the login flow work?”)
  • Make changes (“Change the header color to blue”)

​
Working with the Agent

The agent has five modes, but you don’t need to worry about switching between them—it picks the right one automatically based on your request.

​
Plan Mode

When you ask for something complex, the agent first creates a plan. It shows you what it intends to do before making changes. Review the plan and approve it to continue.

​
Execute Mode

The agent writes code and makes changes to your project. You’ll see files being created or modified in real-time.

​
Debug Mode

When something isn’t working, describe the problem. The agent investigates your code, finds the issue, and fixes it.

​
Build Mode

Ask the agent to prepare your app for publishing. It runs the build process and reports any issues.

​
Chat Mode

Just want to talk? Ask questions about your project, get explanations, or brainstorm ideas without making changes.

​
How to Use Agent Mode

  1. Open your project in OptiDev
  2. Look for the Agent panel on the right side
  3. Type your request in plain English
  4. Watch the agent work and respond to any questions it asks
  5. Review the changes in your project

​
Tips for Better Results

Be specific about what you want
  • Good: “Add a signup form with name, email, and password fields”
  • Vague: “Add a form”
Describe the outcome you need
  • Good: “When users click Submit, save their data and show a success message”
  • Vague: “Make the form work”
Ask follow-up questions The agent remembers your conversation. If something isn’t right, just say “Make the button bigger” or “Change the color to red” without repeating the whole context. Let the agent handle errors If you see an error message, copy and paste it to the agent. It can usually diagnose and fix the problem.

​
Auto-Save and History

Your conversation with the agent is saved automatically. If you close the browser and come back, you can pick up where you left off. The agent also auto-commits your changes to version history, so you can always roll back if needed.

​
For Developers

​
Agent Modes (Technical)

ModeAccess LevelUse Case
planRead-onlyArchitecture planning, code review
executeFull write accessFeature implementation
debugFull write accessError investigation and fixes
buildBuild system accessCompilation, bundling
chatRead-onlyQ&A, explanations

​
Session Persistence

Agent sessions are persisted server-side, reducing token costs by 50-70% for returning conversations. Sessions include full conversation history and project context.

​
MCP Tools Available

The agent has access to Model Context Protocol (MCP) tools including:
  • File system operations
  • Terminal commands
  • Database queries (OptiDev Cloud)
  • Image generation (Gemini 2.5 Flash)
  • Environment variable management

​
WebSocket Events

Agent communication uses WebSocket streaming. Events include:
  • agent:message - Agent response chunks
  • agent:tool_use - Tool execution notifications
  • agent:complete - Stream completion
  • agent:error - Error handling

​
API Reference

POST /api/claude-agent/message
WebSocket: /api/claude-agent/stream
GET /api/claude-agent/session/:id