TreeMap
A free, open-source disk-space visualizer. Squarified treemap, duplicate hunting, disk-full forecasting, and cleanup that only ever moves files to the Trash.
A native Mac app for a roster of AI bots that run on the Claude Code or Codex plan you already pay for. Free for macOS 14 or later.
ProductivitySwift · GoGitHub @xocketlabsOnlineOwner unconfirmed
Have you tried it? One tap is enough. No sign-in needed.
Owners of public repositories under a personal GitHub account can sign in to verify ownership and manage the introduction themselves. No administrator approval is needed.
Sign in, generate a verification file, and commit it to this repository’s default branch within 30 minutes. Files on other branches or forks do not count. Then check the file below to manage the project without administrator approval. Keep the file to retain your access.
If needed, you will go to sign-in. After signing in, return here and press your chosen verification button again.
Run /nomorevibe verify in your project folder to verify ownership and manage the introduction yourself.
Verify in one minuteAsk about other verification options
Native macOS app: a team of AI bots on your own Claude Code or Codex subscription. Source-available (BUSL 1.1), no backend, no telemetry.
Taken from the repository README
Loading the full README. The saved excerpt is shown below.
A Mac app where your AI bots live. They use the subscriptions you already pay for.
Each bot is a real agent: claude or codex running under your own sign-in, on your Mac or
on a machine you rent. Give one a job, a schedule, or a standing goal. They hand work to each other,
and they ask before they act.
Early access. In daily use, not yet Apple-signed. Download the DMG from
Releases, or build from source. See
CHECKLIST.md for exactly what is and is not done.Read the full README ↗A Mac app where your AI bots live. They use the subscriptions you already pay for.
Website · Download · Changelog · Contributing · Security
Each bot is a real agent: claude or codex running under your own sign-in, on your Mac or
on a machine you rent. Give one a job, a schedule, or a standing goal. They hand work to each other,
and they ask before they act.
Early access. In daily use, not yet Apple-signed. Download the DMG from Releases, or build from source. See CHECKLIST.md for exactly what is and is not done.
Requires macOS 14 or later, on Apple silicon or Intel.
Agent products from model vendors are closed by construction: one model, one price floor, your work on their servers. bupsy is the opposite shape.
| Chat | Claude Code, Codex, or a local model in Ollama or LM Studio, switchable per bot, each keeping its own thread. Bubbles, tables, code with a Run button, attachments, replies, voice in and out |
| Teamwork | Bots message each other, hand work on, create new bots, file bots into sections, and change each other's job, model or chat — within the trust you give them |
| Routines | "Brief me every weekday at 7am" — runs with the app closed. Bots can schedule them too |
| Standing goals | "My inbox has nothing older than a day" — it checks, and acts only when something is off |
| Fresh chat | A long chat moves on to a new one from the bot's own hand-off notes, by button or by itself, so each message costs far less |
| Commands | Type / for Bupsy's own commands, your Claude Code commands and skills, or your Codex prompts |
| Your apps | Any MCP server — GitHub, Slack, Notion, files |
| Ask first | Four trust levels, from "ask me everything" to "never ask" |
| Carry on from Claude Code | Pick a conversation from VS Code or the terminal and it becomes a bot that remembers it |
| Your other devices | Use Bupsy from your phone or another Mac over Tailscale; link Macs to see their bots, and move a bot or a whole section to another Mac |
| A computer (beta) | Give a bot a cloud machine so its work survives a closed lid |
bupsy spawns vendor binaries and never holds vendor credentials. Signing in happens in a terminal we render but do not read; the credential lands in files owned by you and the CLI. There is no column in this codebase that could hold one, and a test walks the live database schema to prove it.
The same rule shapes connectors: MCP config references environment variables by name, never by value, so a token never passes through a file we wrote. That is more setup than a paste box would be, and the app says so rather than hiding it.
Read docs/00-compliance-and-auth.md before touching anything that authenticates. It quotes the vendor terms the design is built around.
| Platform | Status |
|---|---|
| Mac (macOS 14+) | The app. Everything happens here. |
| Phone and other Macs | In the browser, through your Mac over Tailscale: Settings › Use from other devices. Linked Macs show each other's bots and can move bots between them. |
| iPhone app | 🚧 In development. Not available to download yet. |
| Windows and Linux | 🚧 In development. Not available to download yet. The Go core (bupsy-run) already runs there. |
Requires macOS 14+, Xcode 16 with Swift 6, and XcodeGen. The headless core needs Go 1.24.
brew install xcodegen
make run # generate the project, build and launch
make test # the BupsyKit tests
make spike # drive your real claude binary and watch the event stream
You will need claude or codex installed and signed in. Bupsy detects whichever you have.
./scripts/ci-local.sh runs what CI runs.
| Path | What it is |
|---|---|
Apps/BupsyMac | The Mac app (SwiftUI) |
Apps/BupsyiOS | The iPhone app, in development |
Packages/BupsyKit | The core: engines, hosts, approval gate, store, remote access |
Tools/bupsy-run | The headless core in Go, for Linux and Windows, plus bupsy-mcp, the tools bots use to talk to Bupsy |
conformance | Shared test cases both cores must agree on |
site, functions | bupsy.si: the landing page, download form and MCP server |
docs | Compliance, decisions, design, release process |
A SwiftUI app wraps BupsyKit, which owns two abstractions:
ExecutionHost — where work runs: this Mac, or a machine you rent.EngineAdapter — one per CLI. Spawns the published binary and normalises its native output
into one AgentEvent type.Everything above that line — chat, transcript, audit log, approval gate — is written once, and every engine and host plugs in beneath it. PLAN.md has the long version.
Issues and pull requests are welcome. Read CONTRIBUTING.md first: a few of its rules (never hold a vendor credential, clean room) are not negotiable. Contributions need the one-time CLA. Everyone follows the Code of Conduct. Report security problems privately, as SECURITY.md describes.
Source-available under the PolyForm Shield License 1.0.0. Licensed by Xocket Labs Inc., a Delaware corporation, part of Xocket Foundry.
Use it free, personally or at work. Read it, change it, build it, share it. What you may not do is use it to make a product that competes with Bupsy: a fork, a clone, a rebranded or hosted version, paid or free. That rule does not expire. This is not an open-source licence by the OSI definition and we do not call it one; docs/02-licence.md explains the choice. Version 0.2.0 and earlier were published under the Business Source License 1.1, which still applies to those versions. Third-party components keep their own licences; see NOTICE.
Not affiliated with, endorsed by, or sponsored by Anthropic, OpenAI, xAI or any other vendor. Product names are their owners' trademarks, used only to say which tools bupsy runs.