Your coding agents in one window

One app instead of a terminal per agent

Coding agents come with their own command-line tools, their own windows and their own habits. AyCode starts their official tools on your computer and gives all of them the same calm surface: one sidebar for every chat, one input field, one way to approve a step, and the same view of what changed.

Each provider runs in its own process. If one goes silent, only its chat waits; the window, the other chats and your typing never freeze. And if an agent returns nothing, the chat says why instead of showing an empty answer.

Which agents work with AyCode

• Anthropic’s Claude Code, with Fable, Opus, Sonnet and the other Claude models.
• Codex through OpenAI’s app server, and local models from Ollama or LM Studio without any sign-in.
• Antigravity with Google’s Gemini models, Grok from xAI, Kimi Code, Mistral Vibe, ZCode for Z.ai GLM, DeepSeek and Hermes.
• OpenCode, Cline, Pi and Oh My Pi, open agents that reach many models, for example through OpenRouter.

AyCode can find, install and test these apps for you, after you agree. Several Codex accounts are fine too, for example a private one and one for work. The provider setup guide explains how to connect each agent.

Models, effort and permissions per chat

Click the model name to search across all providers. Next to it you choose how long the model thinks and, for Claude and Codex, how fast it answers. Three modes decide how freely an agent acts: Normal asks before commands and file changes, Plan writes a plan first and changes nothing until you approve it, and YOLO never asks, for throw-away folders and worktrees.