Single-Tenant AI: Singularity v6 Runs Inside Your Tenant

The Data Residency Constraint

For the last three years, AmLaw firms have watched AI vendors ship "legal copilots" and "document review agents." Most of them live in vendor clouds. For many firms, especially those with strict information-governance policies or regulatory constraints, those tools have been off-limits. Sending active matter files to a multi-tenant SaaS platform, even with contractual promises, doesn't pass audit.

The result: most of the legal AI wave has skipped the firms that could afford to deploy it at scale.

This changes with Singularity v6.

What Singularity v6 Does

Singularity v6 is a platform that runs entirely inside your Azure tenant (or AWS, or your equivalent). The firm's curated knowledge graph - the data layer that grounds every Atlas response - stays on your infrastructure. So does every execution path: document ingestion, agentic curation, document review workflows, and drafting all happen inside your boundary.

The architecture is simple: Atlas connects to your existing systems (iManage, SharePoint, OneDrive, NetDocuments, email) via OAuth or service principals. As new documents land, Atlas agents read, classify, and link them into a per-tenant graph. The curation is continuous and agentic, not a one-time static ingest. That means the graph stays sharp as your matter data evolves.

On top of that graph sit four product surfaces: Workspaces (where agents review hundreds of documents in parallel and draft new client docs from the findings), Lists (reusable agent action sequences teams build once and run on every matter), Agents (the execution layer, both for in-workspace review and for ongoing KG curation), and MCP + API for Claude (so Claude and Claude Code can read your matters and documents directly with the same access controls).

None of this requires moving your data out of your tenant.

How We Built It

Singularity v6 required two technical shifts from earlier Atlas versions.

First, multi-tenancy at the data layer. Every firm gets a dedicated knowledge graph instance, with strict access isolation enforced at query time. When an agent reads documents to curate or review them, it queries only the graph for its tenant and matter.

Second, OAuth and service-principal flows for every integration. This week we shipped app-only Graph consent flows so IT teams can grant Atlas read access to SharePoint and OneDrive without creating shared mailboxes or service accounts. We also shipped a cloud picker that lets teams browse and select specific iManage matter folders, SharePoint sites, and OneDrive locations to seed into the graph. No bulk exports. No manual folder mapping. Pick the source, and the agents start curating.

Who This Is For

This is built for AmLaw firms where information governance isn't a checkbox - it's a veto. Compliance teams, audit, privacy officers. Firms that have spent three years watching AI vendors ship and saying "not until it runs inside our boundary."

It's also built for firms that want to move fast. Single-tenant deployment often meant slow, expensive, on-prem software. Singularity v6 is neither. It's cloud-native software running inside your cloud account. Deployment takes days, not months. Updates roll out automatically. The firm owns the infrastructure spend and the data sovereignty.

What Changes

Before Singularity v6, firms wanting AI-powered document review and drafting faced a choice: ship data to a vendor cloud, or build it in-house at 10x the cost.

With v6, the choice changes. Matter data stays in your tenant. The curated knowledge graph lives on your infrastructure. Agents run your review workflows, draft your client documents, and curate your own legal precedents. Access control is your access control - the same identity layer and permissions you already use for iManage and SharePoint.

If your firm has been blocked on legal AI by information-governance constraints, this is the unlock.

How to Start

Singularity v6 is live now. If your firm uses iManage or SharePoint and has an Azure tenant, you can deploy in days. Your IT team connects Atlas via OAuth (no credentials shared). Your practice teams start staging matter documents into Workspaces. Agents begin curating the graph and running the review work.

The curated knowledge graph becomes the foundation for every Atlas surface: agentic document review at scale, reusable practice-area workflows (Lists), multi-step agent execution (Agents), and programmatic access via Claude's MCP integration.

Matter data in your boundary. AI agents in your workflow. No compromise on either.