AI Agent-Assisted Development

Pro Only

About AI Agents

AI agents are systems that can autonomously plan, execute, and iterate on multi-step software development tasks. If you haven't tried one yet, these are the important things to know:

This new technology changes the role of the software developer. Rather than writing code by hand or by manually invoking an AI for help, the developer operates as more of a software architect and manager of one or more intelligent agents that handle the bulk of implementation work. You define the goals, make design decisions, and review the agent's output before accepting it into your codebase. The AI agent takes care of all the details — reading and understanding existing code, writing new code, writing and running unit tests, debugging failures, and iterating until the task is done. With Wing's Tasks Manager, this goes a step further: the agents work as a managed development team, and you work as its product designer, answering questions and reviewing finished work.

The advent of AI agents also changes what an IDE needs to do. Wing now provides tools not only to the human developer but also to the AI agents doing the work. Some of these tools reduce the cost and time of AI-assisted development — for example, by giving the agent efficient access to test results, debugger state, and code analysis without consuming excessive tokens. Other features help humans create and assign tasks to AI agents, and then review their work.

Wing's AI Agent Integration

Wing runs your choice of AI coding agent directly within the IDE: Claude Code, Codex, Antigravity, Copilot, Cursor, goose, OpenCode, or any other agent that speaks the Agent Client Protocol. Each of these can read and edit files, run commands and search code, and each is given Wing's testing, debugging and code analysis tools through the Model Context Protocol (MCP). Wing extends what the agent can do on its own by providing you with tools to manage planning, execution, and review of its work.

The agents are not equally safe or equally suited to autonomous work. The table in Know Each Agent's Safety Gaps, in AI Agent Safety, compares them and lists what each is missing.

Whichever agent you choose, the workflow is the same one this chapter documents. With it, you can:

Requirements

Each agent is a separate download with its own account or API key, and Wing does not install agents for you. Setting Up an AI Agent lists what each one needs and how to install it.

Free and low-cost plans may be enough for evaluation; sustained day-to-day use may require a paid plan or comparable API spending. For larger amounts of work, subscriptions are generally much cheaper than paying for usage on an API key.

Wing Pro does not include the legacy AI features (AI Coder, AI Chat, and AI Suggestions), which are only in Wing Classic. An AI provider's API key used with them also works with an agent, as described in Setting Up an AI Agent.

For a step-by-step walkthrough of the workflow this chapter documents, see the AI agent track in the Tutorial. To watch agents at work in Wing on three example projects, see the demos on Wingware's website.

Getting Started

To get started, use Set Up for AI Agent in the Project menu and choose the agent you want to use. This authorizes that agent for the project and writes the files it needs to work with Wing. See Setting Up an AI Agent for details, including what each agent requires.

Once set up, the AI Agent tool panel holds your agent sessions. Depending on the agent, this may be a full terminal session running the agent's own interface, or Wing's UI for interacting with agents via ACP. You can ask the agent questions, request code changes, and use it for complex multi-file development tasks. See The AI Agent Tool for details.

The Tasks tool lets you plan goals and tasks, queue them for execution, and run them sequentially through agent sessions. As tasks complete, you review their changes and track what was done. This is useful for preparing multiple work items in advance and having them run with automatic advancement. See The Tasks Tool for details.

When you want to hand over more of the work, the Tasks Manager agent can run the queues for you. It decides what can run concurrently and what can't, gathers all the agents' questions on the Questions tab, files problems found along the way on the Plan tab for you to review and queue, and acts as a mentor that can explain or fix workflow problems. When to start using it is up to you: many people first work with a single agent until they trust its results, while others turn on the Manager from the start. See The Tasks Manager Agent, and Autonomous Development in Using AI Agents Effectively.

Wing also provides quick access to AI agent-driven actions, including one-click fixes for failing tests, debugger exceptions, and code warnings, as well as code actions from the AI actions icon in the top right of the editor for explaining code, reviewing for quality or security risks, and making changes for optimization or documentation. See AI Actions for details.

Preferences and per-project permissions for AI agent behavior can be configured from the AI page in Preferences and the AI Agent tab in Project Properties. See Configuration for details and AI Agent Safety for a broader discussion of safety considerations when working with AI agents.

Section Contents