The Knowledge Graph as Your Firm's Data Layer
Atlas Singularity v6 ingests your firm's existing data from iManage, SharePoint, OneDrive, NetDocuments, and email into a single curated knowledge graph. Atlas agents continuously read, classify, link, and refine that graph as new content lands, so every search, draft, and review is grounded in your firm's own work, not a generic LLM.
Until now, that graph lived inside Atlas. Your team used Atlas Workspaces to run agentic document review at scale, and Atlas Lists to build reusable agent workflows. But the graph itself was an Atlas-only resource.
With MCP and API in v6, that changes. Your firm's knowledge graph is now accessible to Claude, Claude Code, and any tool that speaks the Model Context Protocol. The graph becomes your data layer, not your platform feature.
What Shipped This Week
We recently shipped consumer OAuth integrations for Microsoft 365 and Google Drive, so your firm can now onboard SharePoint, OneDrive, and Google Drive into the graph in parallel with iManage and NetDocuments. We also built out the picker UX so your team can select which folders and documents get ingested into the knowledge graph on day one.
On top of that integration work, we've exposed the graph itself: the Model Context Protocol (MCP) and a public REST API now let Claude and any other MCP-aware client read your matters, documents, clauses, and agent-curated Lists. Access control is enforced at query time, so a user in Claude sees only the documents and Lists they have permission to see inside Atlas.
How Teams Use It
Consider a few workflows:
In Claude Code: Your associate starts drafting an SPA in Claude Code and grounds it in your firm's precedent agreements. Claude queries the knowledge graph via MCP, finds the last five SPAs your firm closed, and uses the actual language as a baseline. The draft isn't generic; it's shaped by your firm's deal history from day one.
In Scripts and Automation: Your legal ops team writes a Python script that queries the API to audit how your firm has handled indemnification language across 50 M&A deals. The script pulls closing documents from the graph, extracts clauses via the same agents that power Workspaces, and produces a single audit report. No manual download, no copy-paste, no inconsistency.
In Terminal and Headless Tools: A partner on your IP team writes a shell alias that queries the graph for recent patent diligence memos. A paralegal pipes a list of prior art documents into a local tool that checks them against your firm's precedent patents. Your firm's knowledge graph becomes as queryable as your file system.
In Claude Itself: A partner pastes a draft license agreement into Claude and asks it to compare against your firm's precedent. Claude queries the knowledge graph via MCP, finds the relevant precedent directly, and produces side-by-side analysis grounded in your actual history.
Access Control and Security
The MCP and API connections inherit Atlas's access control. When a user connects Claude via MCP, the connection uses the same firm workspace credentials they use in Atlas. When they query, the knowledge graph filters results based on their role and matter access. If a user can't see a document in Atlas, they can't see it in Claude either.
The API itself is rate-limited and requires a firm-level API key, so programmatic access is audited and scoped to your organization.
What It Unlocks
Your firm's knowledge graph becomes a shared resource across tools and workflows. You're no longer choosing between "use Atlas for document review" and "use Claude for drafting." You use both, and they read from the same curated source of truth.
This also means you can build custom tools without rebuilding the data layer. If your firm has a specific workflow or tool you want to wire into the knowledge graph, you can do it via API or MCP without Atlas's involvement. The graph is yours.
Getting Started
MCP and API are live in Atlas v6. If you're running Atlas, your admin can generate an API key from the admin panel. Instructions for connecting Claude via MCP are in the docs at https://atlas-ai.io.
If you're not yet on Atlas, start with a demo of how Workspaces and Lists run agentic review and drafting on your firm's actual iManage and SharePoint data. Then wire the graph into the rest of your toolkit.
See it in your environment.
AtlasAI deploys inside your Azure tenant. Private by architecture, not policy.
Request a demo →