Perplexity turns Spaces into collaborative Projects

Every knowledge worker knows the specific pain of starting a brand new AI chat: you paste the same five paragraphs of background context, re-upload the same spreadsheets, and pray the model remembers your formatting rules. When it finishes, it spits out raw text that you have to manually copy back into your actual working docs.

Perplexity wants to kill that loop entirely. The company just rolled out Projects, a major upgrade to its Spaces feature inside Computer, turning what was once a collection of research threads into a persistent collaboration hub where humans and AI agents work directly on the same files.

All existing Spaces are being migrated automatically, and Projects is live for all Computer users starting today.

A shared hard drive where agents edit in place

The most significant architectural shift in Projects is a persistent, hierarchical file system. Rather than generating one-off snippets that live and die inside isolated chat logs, AI agents inside Computer can now open files, read the latest versions, apply edits, and save changes back into the project storage.

That means one team member can log off after generating a preliminary data model, and another colleague can launch a new session the next morning that picks up directly from that saved asset. Scheduled automations can even maintain and update live status reports in the background without anyone having to manually run them.

For desktop users on macOS and Windows, Projects can also hook into local files and folders. To prevent privacy nightmares, those local files stay private to the user who uploaded them and are never exposed to other collaborators in the workspace.

Perplexity also plugged Projects into existing enterprise plumbing. The system connects with over 400 third-party tools, including Google Drive, Notion, Linear, Snowflake, and GitHub. You can also bind Projects directly to Slack or Microsoft Teams channels using the /project command, allowing message threads to instantly feed into project context.

A self-improving brain between tasks

To keep everything from turning into a chaotic digital landfill, Perplexity paired the shared storage with an evolving memory engine dubbed Brain.

Between active tasks, Brain combs through recent project sessions, diffs, and updated documents to maintain a living summary of where the project stands. When a user launches a new prompt, the agent starts from that accumulated understanding instead of tabula rasa.

Granular permissions appear well thought out here. Context is shared across the team, but individual credentials remain strictly siloed. If a teammate queries a connected Snowflake database, the API call executes strictly under that user's personal credentials and permission tier. Users can also fork shared threads to experiment freely without inheriting the original author's private integrations.

Why it matters

For teams wrestling with sprawling initiatives like product specs, investment memos, or cross-functional launches, Projects looks like a massive quality-of-life upgrade. Having an agent that updates your live documents instead of creating yet another duplicate copy removes genuine friction.

Still, whether you should hand it the keys right away depends on your appetite for risk. In-place file editing sounds great until a hallucinating agent rewires a spreadsheet formula or truncates a briefing doc. You will definitely want to lean heavily on project guardrails, like forbidding public actions without manual sign-off, before letting teams run wild.

For the broader AI landscape, Perplexity is executing a clear land grab. It is no longer satisfied being the search engine that surfaces better citations than Google. By embedding persistent memory, multi-player file editing, and direct Slack integration, it is taking dead aim at Notion AI, Slack bots, and Claude Projects simultaneously.

We think this pivot is both defensive and brilliant. Standalone AI query wrappers are constantly at risk of being commoditized by base model upgrades. But once a team anchors its shared documents, ongoing context, and cross-tool workflows inside Perplexity, ripping it out becomes exponentially harder.