On 6 November 2026, Meta introduced Spaces — a feature in the Muse app that lets users describe an idea in chat, while the Muse agent automatically creates a custom interactive space: a gift tracker, a wedding planner, a school events dashboard, or a DIY project board. Spaces are placed on the sidebar or in the main tabs, update in the background through connected apps such as Gmail, Google Calendar, or Notion, and retain context between sessions.
The mechanism is based on describing a task: Muse generates the structure, integrates data from external services, and lets users edit elements directly — marking items as done, adding notes. This differs from ordinary chat in that it turns one-off requests into persistent workspaces where the agent continues working autonomously, tracking changes in prices, dates, or progress without repeated prompts.
Meta's official announcement emphasizes personalization and the convenience of organizing routines, but does not disclose details of the architecture for generating spaces, the frequency of background updates, or limits on connected apps. There are no public benchmarks or ablation studies confirming how much context retention improves compared with previous versions of Muse.
In the landscape of agentic systems, Meta Spaces converges with the approaches of Perplexity Teammates and Anthropic's experiments on integration with external tools, yet differs in its emphasis on user-created mini-apps within a single interface. Unlike specialized enterprise solutions, Meta is betting on mass accessibility and the simplicity of natural-language description, which could accelerate adoption among non-professional users.
The rollout is global wherever Muse is available, including a potential expansion into Asian markets; at the same time, the feature for sharing spaces is still in the research stage and has not launched. This leaves open the question of exactly how collaboration among multiple agents or users within a single space will be implemented — through shared access or state synchronization.
For the industry, this means a shift from episodic interactions to long-term project environments, where the agent acts not merely as an assistant but as a co-author of persistent artifacts. The next steps will likely involve independent verification of the reliability of background updates and the security of shared access.
Key takeaway: Meta Spaces show that sustained collaboration with agentic systems requires not only powerful models but also an architecture of persistent interfaces capable of retaining and updating context without constant user intervention.
