A desktop home for Claude Code.

Why a desktop app for Claude Code

Claude Code is built for the terminal, and the terminal is good at one session at a time. Real work is spread across several projects, each with its own folder, documents, deadlines and history. Switching between them means finding the right folder, restoring the context and remembering what the last session did.

Pandore organizes that work by project. Each project has its folder, its tasks, its conversations and its agent sessions, and Claude Code starts in the right place every time.

What you can do

Built around your projects

• Project context: the agent can use the project's tasks, documents and earlier conversations, so you do not rebuild context in every prompt.
• Sessions that keep running: closing the window does not stop an agent that is working.
• Agent activity by project: one panel shows what is running, what is waiting for you and what has finished, project by project.
• A prompt queue: keep adding requests while an agent works, reorder them, and move an urgent one to the front.
• Clear permissions: choose in Settings whether agents may only read, may also change files, or may run commands, and whether they may search the web.
• Usage you can read: every run is measured by project, client, agent and model in the AI usage dashboard.

Missions: from a goal to finished work

For larger jobs, a Mission starts from an outcome instead of a prompt. Pandore proposes a plan, shows what it will read and write, and waits for your approval. Claude Code then carries out the steps, the result is reviewed, and the deliverables stay in the Mission. Sensitive actions need a precise approval, and a spending cap pauses the Mission when it is reached.

Not only Claude Code

The same projects work with Codex and Gemini CLI, and with 43 local models that run on your computer through the built-in library. You choose Remote, Local or Both in Settings. Missions currently run with Claude Code.