Vicoa: Claude Code & Codex IDE

The signature feature is picking up the exact session you left on your desktop and continuing it on your phone. You type or dictate a reply, approve a pending change, or redirect the agent, and it syncs instantly across phone, web, and desktop. There's no second copy of the session to reconcile. If you want a real Claude Code remote control rather than another cloud editor, this is where it earns its keep. For anyone who has watched a build hang while stuck in a meeting, the point is obvious.

Vicoa is built around running many agents at once, not one at a time. You start Claude Code, Codex, OpenCode, and others side by side, each isolated on its own git worktree, and watch every session's live status in a single list. That makes it easy to spot the agent that's blocked or finished without hunting through terminal tabs. It's closer to an operations console for agents than a chat window, which is exactly what remote AI agent management needs.

Vicoa supports a broad list of agents, including Claude Code, Codex, OpenCode, Gemini, Cursor, Copilot, Kimi, and others, plus third-party ACP-compatible agents you configure yourself. You connect the subscriptions and API keys you already pay for, and Vicoa doesn't add a usage markup on top. Want to switch from Codex to Claude mid-project? You can. That keeps the cost predictable and means you're not locked into one vendor's agent.

Agent output is only trustworthy if you can check it. Vicoa shows each change as an inline git diff with word-level highlights, and you can flip between the changes, the files they touched, and the terminal the agent ran, all without leaving the session. There's also a built-in file explorer next to the chat, so you can browse the tree and drop a snippet straight into your next prompt. No copy-paste shuffle between an editor and a chat box.

A terminal runs right next to the conversation, on the same machine the agent uses. You run commands, tail logs, and watch a build live while the agent keeps working. Because it's the agent's actual environment and not a sandboxed preview, what you see in the terminal is what really happened. If something breaks, you don't have to switch windows to find out why. That's a real difference from a summary you can't verify.

Vicoa includes a Linear-style board with list views, priorities, labels, and sub-tasks. You can right-click a task and start an agent session directly on it. There are also scheduled automations with a run history, so recurring jobs like dependency checks or test runs can fire on their own. Why queue work by hand when the board can hand it to an agent? The free plan caps this at 10 automations; Pro removes the limit.