AI agents

Meet the VibeCaaS AI agents — intelligent assistants that generate code, fix bugs, refactor, and document, working alongside you in the IDE.

AI agents

VibeCaaS AI agents are intelligent assistants that understand your code context, suggest improvements, fix bugs, and write entire features from a description. They work alongside you to accelerate development while keeping code quality high.

This page covers the available agents and how to configure them. For multi-stage automation, see Agent workflows; for getting the best results, see Prompt engineering.

Available agents

Each agent is tuned for a specific job and backed by a model suited to that task.

AgentDefault modelWhat it does
Code AssistantGPT-4Generate code from natural language, complete functions and classes, explain code, suggest optimizations
Debug AssistantClaude 3Identify bugs, suggest fixes with explanations, add error handling, write unit tests
Refactor AgentGemini ProImprove structure, apply design patterns, remove duplication, modernize syntax
Documentation AgentGPT-3.5Write JSDoc/TypeDoc comments, generate READMEs and API docs, draft user guides

Working with agents

Chat interface

Describe what you want in plain language and the agent responds with code plus an explanation.

You: Create a user authentication system with JWT
AI: I'll create a complete JWT authentication system. This will include
    token issuing, password hashing, and protected-route middleware.

// auth.js — generated by the Code Assistant
import jwt from "jsonwebtoken";
import bcrypt from "bcrypt";
 
export async function login(email, password) {
  // ...verify credentials, sign and return a token
}

Command palette

Press Cmd/Ctrl + K anywhere in the editor for quick, in-context AI commands:

  • Cmd/Ctrl + K → "Fix this function"
  • Cmd/Ctrl + K → "Add error handling"
  • Cmd/Ctrl + K → "Convert to TypeScript"

Automated background tasks

Some agents run without being asked:

  • Auto-complete on save — finishes unfinished functions when you save a file.
  • Security scanning — continuously checks for SQL injection, XSS, vulnerable dependencies, and exposed API keys.

Configuration

Customize agent behavior per project from project settings.

{
  "ai": {
    "models": {
      "codeGeneration": "gpt-4",
      "bugDetection": "claude-3"
    },
    "behavior": {
      "codeStyle": "balanced",
      "comments": "detailed"
    },
    "contextWindow": 8000
  }
}
  • Model selection — pick a model per task (code generation, bug detection, and so on).
  • Behavior — set code style (conservative, balanced, creative) and how much commentary to include.
  • Context window — control how much surrounding code the agent can see, in tokens.

Best practices

Writing good prompts

  • Be specific about requirements.
  • Include examples where possible.
  • Specify the language and framework.
  • Mention edge cases to handle.

Reviewing AI output

  • Always review AI-generated code before merging.
  • Test thoroughly before deploying.
  • Check for security issues.
  • Verify business logic is correct.

Next steps