The coding agent

Point exe at Ollama in Windows → Configuration (ollama.base_url and ollama.model; ollama.effort sets the thinking effort on models that support it, or off to disable thinking). A local signed-in Ollama at http://127.0.0.1:11434 can use cloud models like glm-5.2:cloud with no API key; https://ollama.com needs one. Then:

• The Agent tab in a VM window runs the agent inside that VM. It can install packages, write code and start services — it has passwordless sudo inside the VM, and the VM is the sandbox boundary.
• The Chat icon and window appear once a chat backend is usable: a conversation that can see and drive your whole VM cloud. Replies run in the daemon, not in the browser: closing the tab (or losing the network) never interrupts a long task — reopen the chat and select the session, marked with a green dot while streaming, to rejoin it live. The Stop button actually cancels the run.

The Chat window can also run on a ChatGPT subscription instead of Ollama: in Windows → Configuration → OpenAI, click Sign in with ChatGPT… (the OAuth flow the Codex CLI uses — no API key), set chat_provider to openai, pick a model (gpt-5.4, gpt-5.4-codex, …) plus an optional reasoning effort, and Save. The browser sign-in redirects to localhost:1455; the daemon listens on all interfaces there, so when it runs on another machine swap localhost for the daemon's host in that final URL — or paste the URL into the tab's paste field. Tokens live in ~/.exe/openai.json and refresh themselves. While signed in the tab also shows the subscription's rate-limit usage — the rolling 5-hour and weekly windows, with their reset times — and any credit balance. The per-VM Agent tab stays on Ollama.

Prefer your own agent? See Help → Agent Skill Guide: exe serves a /skill.md file that teaches Claude Code, Codex or any other coding agent how to drive the API and the VMs.