Perplexity Computer Projects is the collaboration layer that turns Perplexity's solo AI agent into a shared team workspace — where multiple people work from the same memory, files, and connected tools instead of re-explaining context every time. Launched on July 30, 2026, it scopes persistent memory, file systems, and integrations across hubs and users so a two-person shop gets the same compounding brain as a 50-person team. If your team has been passing AI chat histories back and forth over email or Slack, Projects is the upgrade path — and setting it up takes under an hour.
Last verified: August 2, 2026. Pricing, features, and availability change fast. Check your Perplexity account before deploying to a team.
TL;DR:
- Projects launched July 30, 2026, available to all Perplexity users; Brain (the shared memory engine) requires a Max subscription ($200/month)
- Turns Computer from a single-user agent into a multiplayer agentic workspace with shared files, shared memory, and shared connectors
- Brain's internal benchmarks show 25% better answer correctness on context-dependent tasks, 16% better recall, and 13% lower cost per task
- Perplexity Pro costs $20/month; Max costs $200/month; Enterprise Max costs $325/seat/month
- Biggest risk: shared memory means shared mistakes — designate someone to audit it
What is Perplexity Computer Projects?
Perplexity Computer Projects is a scoping system that lets multiple people share the same AI agent context — memory, files, sessions, and tool connections — inside a shared container called a hub. CEO Aravind Srinivas announced it on July 30, 2026, calling it a "multiplayer agentic operating system for work" with persistent memory, files, and sessions scoped across hubs and users.
Previously, Perplexity had Spaces — persistent research environments where a single user could save threads, upload documents, and set context. Projects replaces and extends Spaces: existing Spaces migrate automatically, and the new layer adds multi-user collaboration, shared memory (via Brain), a shared file system, and per-project connectors and custom skills. One person creates the hub; everyone inside works from the same sources, artifacts, and history.
Projects builds on three existing Perplexity systems:
Perplexity Computer (launched February 25, 2026): A multi-model AI agent that orchestrates 19 frontier models to research, code, build, and manage workflows autonomously. Per the official announcement, it "unifies every current AI capability into a single system" and runs on Claude Opus 4.6 for core reasoning, with sub-agents using Gemini for deep research, Nano Banana for images, Veo 3.1 for video, Grok for speed, and GPT-5.2 for long-context recall. For a hands-on look at what Computer does on a desktop, see our Perplexity Computer for Windows review (2026).
Brain (launched June 18, 2026): A self-improving memory system that builds a "living context graph" from your sessions, connectors, and files. Per Perplexity's own benchmark results, Brain improved answer correctness by 25% on context-dependent tasks, increased recall by 16%, and reduced cost per task by 13%. It launched as a Max-tier research preview and remains Max-only as of August 2026.
SPACE (launched July 15, 2026): A sandbox architecture that runs each task in an isolated Firecracker microVM for security. SPACE separates security concerns across three layers — a Control Plane, Node-level Services, and an in-sandbox Firecracker microVM — so sensitive credentials stay outside the sandbox until the exact moment they are needed. Sessions can be paused, resumed, or branched, with rolling snapshots as frequent as every minute and recovery up to one week later.
Projects extends Brain's self-improvement loop from one person to a whole team. Instead of each teammate maintaining a separate AI context, a hub pools everyone's sessions, files, and corrections into one shared memory graph that compounds over time.
How does Perplexity Computer Projects work?
A hub is the shared container — it can represent your company, a specific team, or an individual project. Inside a hub, three things are scoped collectively:
- Memory: Brain's context graph now belongs to the hub, not to one user. Every teammate's work sessions, corrections, and file interactions feed the same memory. When one person teaches the AI something, every other person benefits. Brain synthesizes the combined activity overnight, building a shared LLM wiki that loads into every session.
- Files: A shared file system acts as a "single source of truth for all collaborators" on a Project. If one teammate uploads the sales pipeline CSV, everyone in the hub can reference it without re-uploading.
- Sessions: Work sessions persist across collaborators. A task one person starts can be continued, iterated, or branched by another, with all context preserved. You can fork a shared thread to build on it without inheriting the original author's private connector access.
This is the shift from instructions to objectives. With Projects, an entire team gives the AI a shared set of objectives, shared files, and a shared memory — and the AI figures out the steps for everyone.
How much does Perplexity Computer Projects cost?
Projects is available to all Perplexity users, but the Brain memory system that powers shared memory requires a Max subscription. Here are the verified pricing tiers as of July 2026:
| Plan | Monthly Price | Annual Price | Projects Access | Brain Access | Key Limits |
|---|---|---|---|---|---|
| Free | $0 | $0 | No | No | Capped daily searches |
| Pro | $20 | $200/year | Basic Projects | No | One-time 4,000 Computer credits |
| Max | $200 | $2,000/year | Full Projects + Brain | Yes | 10,000 Computer credits/month |
| Enterprise Pro | $40/seat | $400/seat/year | Shared spaces | No | File uploads, admin controls |
| Enterprise Max | $325/seat | $3,250/seat/year | Full Projects + Brain | Yes | Audit logging, data retention controls |
Sources: Perplexity pricing page and Perplexity Enterprise Pricing, cross-referenced with third-party pricing guides. Prices verified July 2026.
For a small team of two getting started with shared AI memory: two Max subscriptions cost $400/month. A 5-seat Enterprise Max team costs $1,625/month. The practical question is whether the 25% accuracy lift and 13% cost reduction from Brain justifies the premium — for teams doing repetitive research, coding, or analysis tasks, it likely does. For occasional users, the overnight update cadence may not accumulate enough context to show meaningful benefits. If your team is cost-conscious, start with Pro on one seat, test the Projects file-sharing and connector features, and upgrade to Max only when shared memory becomes the bottleneck.
For a deeper comparison of how Computer fits into the broader tool landscape, see our companion guide on Perplexity Computer app connectors and setup.
How to set up Perplexity Computer Projects for a small team
Setting up Projects for your team takes about an hour. Here is the step-by-step process for a small business or two-person team.
Step 1: Confirm your subscription tier
Log into your Perplexity account and check that you and every teammate who needs shared memory is on a Max plan ($200/month) or Enterprise Max ($325/seat/month). Brain — the self-improving memory system that powers shared context across Projects — is a Max-tier feature. Without it, you get shared files but not the compounding memory.
If only the hub owner has Max, they can create the shared context, but teammates on Pro will not contribute to or benefit from the Brain memory graph. Everyone who should benefit from shared memory needs to be on Max.
Step 2: Create a hub for your business
Inside Perplexity Computer, create a new hub for your organization. A hub can represent your whole company or just one project — for a small team, one company-wide hub with multiple Projects inside it is usually the right starting point.
Name it after your business or the project it represents. Upload your key reference files at this stage: brand guidelines, product documentation, customer databases, SOPs, pricing sheets — anything the AI should permanently "know" about your business. For a detailed walkthrough of connecting external tools, see our Perplexity Computer app connectors guide.
Step 3: Connect your team's tools
Projects carries over Computer's connector system. Connect your hub to the tools your team uses daily:
- Google Workspace — pull docs, sheets, and shared drives into the hub's context
- Slack or Microsoft Teams — let the AI reference team conversations and channel history
- Custom skills — define reusable skills that tell Perplexity exactly how to respond with preferred formats, structure, and steps, so every answer works the way your team expects
Per Perplexity's Projects FAQ, Projects connect to Drive, Slack, Notion, GitHub, Linear, Snowflake, and 400+ other tools, scoped to the sources that matter for the work. Once a hub owner connects a tool, every collaborator in the hub can pull from that same connection. One person connects the CRM; everyone in the hub can query sales data without separate logins.
Step 4: Invite your teammates
Add collaborators to the hub with their Perplexity account emails. Each person gets access to the shared files, memory, and connectors. The key advantage: nobody has to re-explain the business to the AI because Brain already has the context from every prior session in the hub.
Per Perplexity, each person's connected data stays scoped to their own access — collaboration does not pool one teammate's private connectors into a shared source. Forking a shared thread lets a teammate build on it without inheriting the original author's access.
Step 5: Set up a memory audit cadence
This is the step most teams skip — and it is the one that matters most. Shared memory across a team means shared mistakes. If Brain learns something wrong early on, that wrong information can spread to everyone in the hub.
Per Perplexity's announcement, corrections get remembered too — Brain learns from user corrections and dead-end sources. But for the first two to three weeks, designate one person to check the hub's memory graph weekly — look for entries that are outdated, incorrect, or no longer relevant, and remove them. After the initial break-in period, audit monthly.
This is the same principle behind building an AI agent control plane: agents without governance become shadow IT. Brain's shared memory is powerful precisely because it compounds — but that means errors compound too, unless someone is watching.
What are the security and privacy implications of shared AI memory?
Perplexity's SPACE sandbox architecture addresses the execution-security side — every task runs inside an isolated Firecracker microVM that self-destructs when the task finishes. Credentials are injected only at the moment they are needed and never persist inside the sandbox. SPACE supports bring-your-own encryption keys, on-premises deployment, and fully offline operation for sensitive environments. Enterprise administrators get new security controls to configure internet access in Computer sandboxes.
But Projects adds a new privacy dimension: shared memory. Three questions to address before deploying Projects to a team:
Who can see what: Everyone in a hub shares the same memory graph. If one person works on confidential client files, that context is visible to all collaborators. Segment hubs by sensitivity level — do not mix client-confidential work with general team work in the same hub. Per Perplexity, each person's connected data stays scoped to their own access, so collaboration does not automatically pool private connectors into one shared source. But the memory graph itself is shared.
What happens when someone leaves: Brain's memory is source-traceable — every entry links back to the session, file, or connector it came from. But removing a collaborator from a hub does not automatically strip their contributed context from the shared graph. You may need to manually remove their contributions, especially for sensitive data. Build offboarding into your hub management process from day one.
Enterprise controls: Enterprise Max plans add data retention controls, connector access management, and audit logging — the governance features enterprise teams need before deploying persistent shared memory at scale.
For a broader look at how AI coworking tools handle this tradeoff, see our comparison of Claude Cowork and ChatGPT Work as AI coworkers.
How does Perplexity Computer Projects compare to team AI features in Claude and ChatGPT?
Perplexity is not the first AI platform to ship team-scoped context. Here is how Projects compares:
| Feature | Perplexity Projects | Claude Cowork | ChatGPT Enterprise |
|---|---|---|---|
| Shared workspace | Yes — hubs with files, memory, sessions | Yes — shared agent sessions | Yes — shared workspace |
| Persistent team memory | Yes — Brain's self-improving context graph, nightly offline synthesis | Limited — session-shared context | Limited — custom instructions, no compounding graph |
| Shared file system | Yes — single source of truth for all collaborators | Shared documents | Shared workspace with file uploads |
| Tool integrations | 400+ connectors (Drive, Slack, Notion, GitHub, Linear, Snowflake) | MCP connections | Limited integrations |
| Self-improvement loop | Yes — Brain runs offline compute between sessions to improve context | No | No |
| Sandbox isolation | Yes — SPACE / Firecracker microVMs | Container-level | Container-level |
| Pricing (top tier) | $200/mo (Max) / $325/seat (Enterprise Max) | Custom | Custom |
The differentiator is not "teams can use this together" — Claude and ChatGPT both cleared that bar. It is the self-improving shared memory that compounds across a team's activity automatically, without anyone manually curating what gets remembered. Most competing team features handle shared documents and shared chat history well; fewer claim an offline consolidation process that actively gets better at a specific project's context over time, the way Brain's nightly loop does. Whether that compounding advantage holds up under real team usage with noisier inputs is an open question, but the architecture is genuinely novel.
What does Perplexity Computer Projects mean for different team sizes?
Solo business owner
You stop re-explaining your business to an AI every time you open it. Upload your key files once, let Brain build the context graph, and your agent starts every session already knowing your products, customers, and processes. The AI goes from a tool you feed every morning to a colleague who shows up prepared.
Small team (2-10 people)
Your team stops working in silos. Everyone touches the same files, the same research, and the same history. One person's research becomes everyone's research. The cost benefit compounds: at 13% cheaper repeat tasks and 25% fewer corrections, a 5-person team running Computer daily saves meaningful hours per week. For a broader look at how agent operating systems enable this kind of shared context at scale, see our guide on mastering your AI agent operating system.
Team manager
You can finally see what your team and AI tools have been working on in one shared place instead of chasing five different chat histories. Projects becomes the audit trail — the shared memory graph shows what was asked, what was corrected, and what files were touched across the team.
First-time AI user
You do not have to keep repeating yourself. Set the context once and it sticks. Projects is actually a strong starting point because the hub structure forces you to organize your context upfront — which is better than starting with ad-hoc chats that accumulate without structure.
What this means for you
If you are running a business — even a two-person operation — and you are still passing AI chat histories back and forth or re-explaining the same context every session, you are paying for the same work twice. Projects eliminates that waste by making the AI's memory a shared asset.
The businesses that figure out shared AI systems first are the ones pulling ahead. The setup is not complex, but it does require deliberate choices: which tier to subscribe to, how to segment hubs by sensitivity, who audits the memory, and which connectors to wire up first. Spend an afternoon on setup and you compound the savings every day after.
FAQ
Q: Is Perplexity Computer Projects free? A: Projects is available to all Perplexity users, but the Brain shared memory system that makes it powerful requires a Max subscription at $200/month (or Enterprise Max at $325/seat/month). Without Max, you can share files in a hub but will not get the compounding team memory.
Q: What is the difference between Perplexity Spaces and Projects? A: Spaces (launched earlier in 2026) were project-based workspaces for saving related searches, files, and custom instructions within a single user's scope. Projects (launched July 30, 2026) extend the concept to multiplayer collaboration with shared memory, shared file systems, and tool integrations scoped across hubs and users. Existing Spaces migrate to the new Projects experience automatically.
Q: How many people can share a Perplexity hub? A: There is no published hard limit on hub membership. Hubs can represent a company, a team, or a single project. For Enterprise Max at $325/seat/month, teams get additional data retention controls, connector access management, and audit logging.
Q: Can I remove a teammate's contributions from shared memory? A: Brain's memory is source-traceable — every memory entry links back to the specific session, file, or connector it came from, and can be removed individually. However, removing a collaborator from a hub does not automatically strip their historical contributions from the shared graph. You should manually review and remove sensitive contributions during offboarding.
Q: Does shared memory mean shared mistakes? A: Yes — if Brain learns something incorrect early on, that wrong information can spread to everyone in the hub. Per Perplexity, corrections are also remembered and the system should improve as people fix errors. Designate one person to audit the hub's memory weekly during the first 2-3 weeks, then monthly after that.
Q: What tools can I connect to a Perplexity hub? A: Projects supports connections to Google Drive, Slack, Notion, GitHub, Linear, Datadog, Snowflake, and 400+ other tools, scoped to the sources that matter for the work. Mac users can also add local files and folders through Personal Computer. The connector system existed before Projects — the new feature is that one person's connection is shared across the entire hub.
Q: How is this different from what Claude Cowork or ChatGPT Work offer? A: Claude Cowork and ChatGPT Work both support shared agent sessions and shared documents. The differentiator is Brain's self-improving shared memory — an overnight consolidation process that actively gets better at a specific project's context over time. Most competing team features handle shared documents and chat history well; fewer claim a compounding memory loop that improves without manual curation.
Discussion
0 comments