Features

Keep several projects side by side on the same canvas, each with its own tasks and agents. Switch context by moving across the canvas instead of between windows.

Every task you give an agent gets its own card with its agent, mode, status, and context. See at a glance what's running, what's done, and what needs you.

Choose the provider, model, and effort level for each task. Use a heavier model for architecture and a faster one for routine fixes.

Each card keeps its own conversation with the agent, with full history. Come back to a task days later and pick up where you left off.

Talk through the approach with the agent in Plan mode before any code changes. Approve the plan or switch modes with Shift+Tab when you're ready for it to start.

Open any file the agent mentions in chat without leaving Buzzpot. Preview it, read the source, and make quick fixes by hand.

Build a library of prompts you use often, like review checklists, coding conventions, and standard instructions, and reuse them in any task.

Queue a message to send later. It sits in the chat until you send, edit, or delete it. On a card with dependencies, it becomes the prompt that starts the task.

Track rate limits, usage, and how full each agent's context window is, so you know when to start fresh before quality drops.

Drop stickers anywhere on the canvas to leave notes and reminders. Group cards under a name and color to label areas of work.

When an agent spins up subagents, they show up under its card with their own status and context. Open any of them to read what it's doing.

Answer an agent's questions with one click or your own words, approve what it asks to do, and run its slash commands right from the chat.

Set the provider, model, and effort level new cards start with, so every new task is ready to go.

Connect cards to set dependencies and queue a message on each next card. Buzzpot works out what can run in parallel and what has to wait, and starts each task as soon as everything before it is done. You start the first one.

Agents can create cards, connect them, and reshape the workflow through MCP. Ask an agent to break a feature into tasks, and it lays out the plan on the canvas for you to review.

Send another message at any time. Buzzpot stops the current turn and starts a new one on the same card, with the full conversation still there.

Turn them on per project, and Buzzpot handles the Git side of parallel work. Each working task gets its own branch and copy, merges back when the agent reports it done, and cleans up afterward.

When a merge conflicts, that card's agent gets one try to resolve it in its own copy, or you can turn that off in settings. If it can't, the card stops. Cards that depend on it wait; the rest of the workflow keeps going.

Isolated workspaces are off by default: agents work directly in your project folders and you manage branches yourself. Folders that aren't Git repositories work too.

Open your projects, agents, and conversations from your phone or any other device. Check progress, answer agents, and start the next task while you're away from your desk.

Get a macOS notification or a push on your phone when a card is done or an agent is waiting for you.

Your code and CLI agents keep running on your main computer. Other devices only control them, so your project stays where it is. Keep your Mac on while you're away.

Close Buzzpot, open it again, and everything is where you left it: the canvas, cards, conversations, and run state.

If Buzzpot or your computer restarts mid-run, unfinished work is recovered, so you can resume instead of starting over.

Buzzpot backs up its workspace database every day and keeps the last ten copies.