The product

One loop, four columns —
everything in its place.

Every screen below is the real app, running on a real vault. Work flows left to right: tasks, folders, documents, agents — and it always returns to you.

The whole loop on one screen

Tasks on the left rail with live status dots. The folders and references each task works with. The document you're reading. The agent doing the work. Nothing hides behind a window switch.

Rondo workspace: task rail, working folders, a markdown document with a status table, and a running agent side by side

Tasks are first-class — a task is a unit of intent, not a folder. Each carries its own agents and reference folders.
Status at a glance — coral means playing, amber means waiting for you. The rail never lies.
Documents render live — tables, checklists and headings, right next to the agent that wrote them.

Fermata — the app knows when it's your turn

Agents pause for permissions and decisions. Rondo turns that moment into a first-class state: the banner turns amber, the mascot asks for you, a native macOS notification fires. You can walk away — nothing waits silently.

Rondo in the Fermata state: an amber banner reads 'Waiting — please respond to the permission request' and the Dot mascot says 'I need your input'

A state language, not a spinner — Rest · Playing · Fermata 𝄐 · Coda · Off-key. Every session, every dot, every notification speaks it.
Native notifications + floating HUD — see every agent's state above any app, click to jump straight to the waiting chat.
Dot keeps watch — the little guy dances while agents play and taps you when one needs an answer.

Dot mascot with a speech bubble: I need your input — waiting for your answer

Claude Code & Codex, native

Not a wrapper — the real CLI, wearing a chat interface. Every tool call becomes a card. Sessions resume where they left off. Flip to the raw terminal whenever you want.

Tool calls as cards — Read, Write, Bash: each one visible, with outputs a click away.
Real sessions, resumed — close the app, come back tomorrow, continue the same conversation.
You set the leash — five permission modes per session, from confirm-everything to full auto.
Background agents stay visible — output keeps flowing into the chat while you work elsewhere.

Agent chat column: a running Claude Code session with Read and Write tool cards and a live activity bar

Context on three axes

This is why the tagline says work becomes context. Open a vault and your agents inherit everything — no re-explaining, every session.

Memory panel showing the CLAUDE.md cascade: Global, Vault, folder and project levels

Vertical — the CLAUDE.md cascade from global to folder, viewable and editable in one panel. Add a memory with one line.
Horizontal — attach reference folders per task; they're injected into the agent with --add-dir.
Semantic — a structure profile tells agents what your folders mean. PARA is just a preset; a built-in vault MCP server answers structure queries.

Documents live next to the work

Markdown with live preview, editable tables and checklists. Canvases for thinking in space. PDFs and office files. All of it in tabs beside the agent — and all of it plain files on your disk.

A JSON canvas open in Rondo showing a release-day flow diagram beside a running agent

Files are the truth — Rondo's metadata lives only in .aiworkspace/. Delete the app; your vault is untouched.
Obsidian-friendly — wikilinks, callouts, canvases and frontmatter render the way you expect.
Agent output opens itself — a Write card links straight to the new file in a document tab.

Hear the whole piece for yourself.

Free during beta. macOS on Apple Silicon.

Download Rondo v0.0.49