Team collaboration
One goal split across roles, every status visible; deliverables land in the same workspace.
Split AI into a row of roles: lead, engineer, ops, support. Each one has its own desk, its own context and its own deliverables — you name the goal, they pass it along.
Break ship-new-help-center into 6 shippable steps
Rewrite /help route, pass three regression cases
Two homepage drafts — pick one to publish
Cluster this week's 24 tickets into 5 issues
Six capabilities you can use today, all in one workspace.
One goal split across roles, every status visible; deliverables land in the same workspace.
Finished work stays where it is; next time you continue instead of re-explaining context.
Files, browser, databases, internal systems — plug in and go; install skill packs from the market.
Install it on your own machine: it is there at boot, and keeps working when the window is closed.
Use your own OpenAI / Claude / Gemini keys or local models — you own the bill and the quota.
One command on one server brings up the stack; data never leaves your network.
From your own laptop to a whole company on private infrastructure — pick one to start.
The desktop app is free to download; subscriptions unlock cloud duty, quota and team features.
One person, one desk
More work, more quota
Yearly and half-year billing switch on the checkout page; perpetual licences live there too. Prices and entitlements follow the checkout page.
Four claims you can verify: where data lives, whether you can take it, whether it runs on your own iron, and whether you can read the code.
Work data lands in a local SQLite file — nothing queued in someone else's cloud; offline still works.
Export is a first-class feature: one command takes the whole workspace with you.
Docker Compose brings up the stack for your intranet; no public account required.
Apache-2.0, third-party attributions in NOTICE, supply chain you can inspect.
A chat app gives you one conversation. MengBot gives you a team with desks and deliverables: goal breakdown, parallel execution, results kept in the workspace.
BYO: bring your API key or point it at a local model. A trial quota lets you finish the first flow without one.
In a local database on your machine by default; with Cloud it lives on your cloud desktop, still exportable as a whole.
The desktop app is free and unbounded locally; after the trial quota you move to a subscription — exported data stays yours either way.
Windows, macOS and Linux all have a desktop app; self-hosting is Docker, so any machine that runs containers works.
Yes — one docker compose up -d brings up the stack, deploy it inside your network, data never leaves.
Subscriptions bill monthly and include model quota plus cloud duty; a perpetual licence covers the desktop app only.
Request an invoice on the checkout page; unused subscriptions are refunded under the checkout page rules.
Download the installer and let the first role deliver something within fifteen minutes. Not for you? Uninstall.
Questions, feature requests, or just curious how others use it — scan a code and join. Plugin updates and usage questions are announced there.
The WeChat group QR refreshes every 7 days; grab the latest one from the README on GitHub.

