Skip to main content

The Problem

Team knowledge lives in fragments: docs in five folders, decisions in chat threads, domain expertise in one senior colleague’s head. New members ramp up slowly; when someone leaves, context leaves with them. Classic wikis rot because nobody maintains them.

What MemoryLake Changes

A team space turns collective activity into organizational memory that maintains itself. Documents get parsed and indexed as they’re added; conversations with the assistant distill into durable facts; contradictions surface as conflicts to resolve rather than silently coexisting. Every member’s AI assistant answers with the team’s full accumulated context — governed by roles and permissions.

Walkthrough

1

Create the team and invite members

Create a team in the console and invite colleagues by email or invite link (both inviter and invitee need verified email addresses). Assign roles: Owner, Admin, Member — or define custom roles for finer control.
2

Structure workspaces around how you work

Create a workspace per major domain — “Product”, “Customer Success”, “Legal” — or per client engagement. Members join workspaces automatically; each workspace holds its own projects, memories, and agents.
3

Load the knowledge

In each project, upload the documents that matter: specs, contracts, research, meeting notes. Folder uploads preserve structure; PDF, Office, Markdown, and images are all parsed and indexed. Existing knowledge in personal notes? The migration tooling in the agent plugin can import prior conversations and memories.
4

Govern access with roles and resource scopes

Beyond the three base roles, MemoryLake supports per-resource scopes: grant a contractor access to workspace A’s “Vendor audit” project and nothing else. The Permission Reference lists every fine-grained capability — from project:doc_read to project:mem_conflict_resolve.
5

Let organizational memory accumulate

As members chat with the assistant and agents work in project context, MemoryLake extracts shared facts: “We committed to SOC 2 by Q2”, “Customer X’s contract renews March 15”, “The refund window is 30 days.” Everything carries a source trace — click any memory to see the conversation or document it came from.
6

Resolve conflicts, keep truth current

When the pricing changes and someone tells the assistant the new numbers, the old memory doesn’t linger silently — a conflict is raised. Whoever holds the mem_conflict_resolve permission reviews and resolves it. Organizational memory stays trustworthy because disagreement is visible.
7

Manage quotas and watch usage

Allocate per-member quotas so costs stay predictable, and use the team statistics page (member rankings, consumption trends, model usage) to see how the team actually uses AI.

What This Feels Like Day to Day

  • A new hire asks the assistant “How do we handle enterprise refunds?” and gets the current policy with a citation — no scavenger hunt, no bothering the one person who knows.
  • An account manager preps for a renewal call: “Summarize everything about Customer X” pulls contract terms, past escalations, and the promise made in last quarter’s QBR.
  • The senior engineer leaves; their design rationale — extracted from months of project conversations — stays searchable.

Governance Highlights

Go Deeper

Team Collaboration

Roles, invitations, quotas, and analytics in detail.

Permission Reference

Every fine-grained permission, explained.