Cycling through all six. Tap any point to stop.

Measured on six things

Free Mac app for a team of AI bots that run Claude Code or Codex on the plan you already pay for.

Security

The architecture is the security story: no Bupsy account, no vendor server, no stored credentials, no telemetry, local files only. But the domain itself scores 4/12 on headers — no HSTS, CSP, or DMARC — and there's no SOC 2, pen-test, or bug bounty published.

Screenshots

1 / 1

Bupsy

About this app

Bupsy is a native Mac app where a small team of AI bots lives and works for you. Each bot runs Claude Code or Codex under your own sign-in and on the subscription you already pay for. There is no Bupsy account, no credits to buy and no Bupsy server; conversations are files on your own disk, and there is no telemetry. Make a bot, tell it what it is for, and it gets on with it: on a schedule, when files change, or when another bot hands it work. Anything beyond reading stops and asks, with four trust levels per bot and an audit log. Free, macOS 14 or later. Not affiliated with Anthropic or OpenAI.

Key features

1Team of AI bots running Claude Code or Codex under your own sign-in and existing subscription — no credits, no Bupsy account
2Six bot faces as status (Ready, Working, Needs you, Done, Asleep, Stuck); routines, goals, and multi-bot handoffs with @mentions
3Approval-gated actions with four trust levels per bot, audit log, local file storage, no telemetry, no vendor server

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

Mac-based software developers

Pain points

Repetitive coding tasks, context switching, manual CI/CD fixes

Primary needs

Automate routine dev tasks with AI while keeping full control and using existing Claude Code subscription

Demographic

IT operations and DevOps teams

Pain points

Manual monitoring, alert fatigue, slow incident response

Primary needs

Schedule bots to check systems and handle routine ops tasks with approval workflows

Demographic

Productivity-focused professionals and small teams

Pain points

Too many tools, scattered information, lack of automation

Primary needs

A team of AI assistants that can connect to Gmail, Notion, etc., and run routines without extra cost

Scout Review

Written by AI from measured evidence, scored out of 100.

Bupsy is a native Mac app that turns your existing Claude Code or Codex subscription into a small team of AI bots with faces, routines, and approval gates. The design is genuinely charming — status as bot expressions, a Finder-like sidebar, a bot creator with live preview. Speed is excellent (93 mobile, 100 desktop Lighthouse). Security is architecturally strong (no server, no credentials, no telemetry) but the domain's headers are weak and no audits are published. Accessibility is solid at 96/100 with one contrast miss. Growth is a sharp wedge aimed at a narrow but real audience.

Measure by measure

Usability

Onboarding is the pitch: no account, no credits, sign in to the CLI you already pay for. Approval prompts are unmissable with explicit Allow/Deny. Docs depth beyond the homepage is thin, and the iPhone/Windows/Linux builds are still 'in development'.

Design

A genuinely characterful Mac app: six bot faces replace spinners as status, the sidebar reads like Finder, and the illustration system holds together across hero, roster, and app previews. Docked for a low-contrast hero input and a stray quota badge.

Speed

Lighthouse mobile 93/100 (LCP 2.6s, CLS 0.001, TBT 0ms) and desktop a clean 100/100. The marketing site is light, static-feeling, and snappy; the native Swift app sidesteps the usual Electron bloat entirely.

Security

Accessibility

Lighthouse accessibility 96/100 with a single failing audit — insufficient color contrast, matching the low-contrast hero placeholder seen in the visual audit. Keyboard-friendly native Mac UI is a plus, but there's no WCAG statement and the site is English-only.

Growth

Sharp wedge: your existing Claude Code/Codex subscription, no credits, no vendor server, local files, approval-gated automation. Specific Mac-dev/DevOps audience with a real pain. Ceiling is the Mac-only, power-user niche until the phone and Windows builds ship.

Improvement areas

Domain security headers: no HSTS, CSP, or DMARC; 4/12 on the HTTP header checklist
Accessibility: one failing contrast audit on the hero input; no WCAG statement; English-only
Platform coverage: iPhone, Windows, and Linux builds listed as 'in development' but not downloadable

Design & Usability

Bupsy's design is the rare case where the product's personality is the interface. Six bot faces — Ready, Working, Needs you, Done, Asleep, Stuck — replace spinners and dashboards, and the macOS sidebar reads like Finder with pinned bots and category sections. The bot creator gives instant visual feedback on shape, color, and mood. Usability follows the same philosophy: no account, no credits, sign in to the CLI you already pay for, and approval prompts are unmissable with explicit Allow/Deny. The gaps are small but real: a low-contrast hero input, a stray quota badge, and documentation that doesn't go much deeper than the homepage.

Speed & Security

Speed is a genuine strength. Lighthouse mobile scores 93/100 with LCP 2.6s, CLS 0.001, and TBT 0ms; desktop is a clean 100/100. The native Swift app sidesteps Electron entirely, so the snappiness isn't just a marketing page trick. Security is more interesting: the architecture is the argument — no Bupsy account, no vendor server, no stored credentials, no telemetry, conversations as local files. That's a strong privacy posture. But the domain itself scores 4/12 on HTTP headers, with no HSTS, CSP, or DMARC, and there's no SOC 2, pen-test, or bug bounty published. The app is trustworthy by design; the website hasn't caught up.

Accessibility & Growth

Accessibility lands at 78: Lighthouse scores 96/100 with one failing audit — insufficient color contrast — which matches the low-contrast hero placeholder seen in the visual audit. There's no WCAG statement and the site is English-only, but the native Mac UI is keyboard-friendly by default. Growth is the more interesting question. The wedge is specific and real: run Claude Code or Codex on the subscription you already pay for, with no credits and no vendor server. The multi-bot handoff, routines, and approval-gated automation target Mac developers and DevOps with a genuine pain. The ceiling is the Mac-only, power-user niche until the iPhone and Windows builds ship — and the domain is brand new, so this is a bet on execution, not traction.

Conclusion

Bupsy is one of the more coherent indie Mac apps in this space: the pitch, the interface, and the architecture all point the same direction. If you already pay for Claude Code or Codex and you've been annoyed by credit-based AI tools, this is the obvious thing to try — it's free, it's native, and it doesn't ask you to trust a server. The work ahead is unglamorous: fix the contrast, add HSTS and DMARC, publish a security page, and ship the iPhone and Windows builds before the Mac-only niche caps the ceiling.

Suggestions for improvement

Fix the hero email placeholder contrast — it's the one thing Lighthouse accessibility flags.
Add HSTS, CSP, and DMARC to the domain; a security-first product shouldn't score 4/12 on headers.
Publish a security page documenting the local-only architecture, threat model, and any audits.
Ship the iPhone and Windows/Linux builds or move them off the homepage until they're downloadable.

How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

	Bupsy	OpenAI ChatGPT	Anthropic Claude	GitHub Copilot
Runs on your existing AI subscription (no separate credits)	Yes — Claude Code or Codex under your own sign-in	No — requires separate ChatGPT subscription	No — separate Claude subscription or API billing	No — separate Copilot subscription
Native Mac app with local file storage	Yes — native Swift, conversations as local files, no telemetry	No — web/cloud, data on OpenAI servers	No — web/cloud, data on Anthropic servers	Partial — IDE plugin, cloud-backed
Scheduled routines and multi-bot handoffs	Yes — routines, goals, @mention handoffs, lead bot can spawn teammates	No — single chat, plugins only	No — single chat interface	No — inline code completion only
Approval gates and audit log for risky actions	Yes — four trust levels per bot, shell commands never auto-allowed, audit log	No — no per-action approval workflow	Partial — tool-use confirmations in Claude Code, not a roster	No — suggestions only, no action execution
Free to use	Yes — free, macOS 14+, source on GitHub	Freemium — paid tiers for GPT-4 class models	Freemium — paid tiers for higher usage	Paid — free tier limited