Moxt

Moxt gives every user a personal assistant named momo that records what it learns about you in plain files. Say once that you prefer concise reports in a table, and that preference carries into future conversations. Most tools forget everything the moment you close the tab. This one doesn't. That's what makes the workspace feel less like a search box and more like a colleague.

Beyond momo, you can build additional AI teammates, each with a name, a role, and its own rules. A social-media brand builder, for example, might run a design lead, a researcher, and a contrarian reviewer side by side, each one responsible for a different slice of the work. You can also pull presets from a marketplace covering common roles. Every teammate is independent, which means one agent's finished work becomes context the next one can act on. That's what makes it a true multi-agent workspace. Nothing falls through the cracks.

Documents live in directories as Markdown, spreadsheets as CSV, and visual reports as HTML. Agents can browse, search, and read these files the way a person opens a folder, without decoding compressed Office formats first. Word, PDF, and Notion imports get converted in the background. It looks simple. It works. If you've built skills in Claude Code or similar tools, many of them drop into Moxt unchanged.

You can attach a timer to a teammate, so it runs on its own. A researcher set to scan the news each day at 13:00 will deliver a summary without anyone asking. This turns Moxt from a place you visit into a service that works in the background. Set it once. Forget it. Missed work is easier to catch too, since each agent's output stays visible in the shared space.

AI teammates plug into Slack and Feishu, so you can call them from the chat window you already use. Instead of opening a separate app, you mention the agent and hand off the task in place. For teams that live in a chat tool, this removes a real chunk of copy-pasting between platforms.

Moxt doesn't sell seats. Members are free, and you pay for the models your agents consume at the provider's published rate, with Moxt adding no markup of any kind on top of what the model vendor charges. A GPT-5.5 call is billed at OpenAI's price, for instance. Credits run at $1 per 100 and don't expire. It's a straightforward model if you like knowing exactly where the money goes.

Agents need a place to run code, and Moxt offers an isolated cloud sandbox per agent, free for now and billed by active runtime only. There's also a local option in the works that lets agents use your own machine. For teams that don't want to manage infrastructure, the cloud path means an agent can start working with zero setup. No servers to spin up.