Fractal: an early-access workspace for shared AI agents and reusable projects
A team workspace built around shared context
Fractal presents a workspace where people and AI agents collaborate on tasks, information and deliverables. Its website describes agents that remember preferences, share context and coordinate work across a project, with visible responsibilities and review points.
The examples include planning a campaign, drafting content and preparing a website. The publisher explicitly labels the demonstrations as illustrative examples with fictional data and local interactions. They show the proposed experience rather than evidence of a real company's campaign results or Franklin's own use of the product.
Give the work somewhere to live
The site describes organized libraries for clients, products and content, with views such as calendars, boards and dashboards. Agents and people are intended to work from the same information instead of reconstructing context in each conversation.
Fractal also presents websites, reports, dashboards and small applications as possible deliverables, hosted in the workspace and reviewed before sharing. Connected channels listed include WhatsApp, Slack, Telegram and Discord; email is described as coming soon. The FAQ says each channel requires its own setup and permissions, so a channel name should not be read as an automatically configured integration.
Reuse a workflow, subject to access
A plugin can package agents, instructions, tools, collections, views and deliverables for another project. That makes the reusable unit larger than a single prompt: it includes both the working context and the tools used to produce an outcome.
Fractal is preparing early access by invitation. The current call to action is joining a waitlist, and the site says access details arrive with an invitation. It should therefore be treated as a product to follow or evaluate when access becomes available, rather than an immediately available workspace Franklin has tested.
Franklin's suggested evaluation is one repeatable project with a clear reviewer and a small set of permitted tools. Inspect whether the retained context and packaged workflow actually reduce setup work while keeping approval decisions visible.