ACL-aware retrieval
Retrieval honors the permissions of your source systems on every request. Nobody sees a document, clause, or matter they could not already open.
One hard trust boundary. Reframe is the single governed layer across it, Microsoft-native, single tenant by construction. Your data never leaves your boundary, and every answer is grounded while every action is audited.
We stand up a hard trust boundary inside your tenant. Your keys, your network, access control enforced at query time. One boundary, and Reframe is the single governed layer across it.
A dedicated environment per firm. No shared compute, no shared index, no co-mingled data. One tenant, one boundary.
Customer-managed keys and your own network controls. Reframe runs where your data already lives, under your Azure governance.
Permissions are evaluated on every request against your source-system ACLs. Answers only ever reflect what the asker may already see.
Built on the Microsoft cloud you already run, with Reframe as the single governed control plane across every model, tool, and action.
Every Reframe deployment is a dedicated, single tenant environment stood up inside your own Azure tenant. The boundary is dedicated to your firm, and everything below holds on day one.
One enclave for one firm, provisioned inside your Azure tenant. No shared compute, no shared index, no co-mingled data.
Encryption keys stay in your hands, in your own key vault, under your rotation and revocation policies. Reframe operates under the access you grant.
The enclave runs on private endpoints inside your network perimeter. No public data plane, and no traffic your firewalls and monitoring cannot see.
Permissions are evaluated on each request against your source-system ACLs, reads and writes alike. Nobody, human or agent, touches what they could not already open.
Every model call, tool invocation, and agent step is recorded end to end, so any answer or action traces back to its source, its actor, and its approval.
Ingestion, indexing, storage, and inference all run inside the enclave. Your knowledge never crosses the boundary, at rest or in flight.
Inference runs against model endpoints deployed within the boundary. Prompts, context, and outputs never transit a shared or external service.
Your data is never used to train, tune, or evaluate shared models. What the firm knows stays the firm's alone.
Every component that can touch firm data runs inside the boundary, under the same keys, the same network, and the same audit trail as the rest of the deployment.
The enclave is built to the frameworks your firm is measured against, so the controls are evidence, not a promise.
Governance is not a policy binder. It is enforced in the path of every query and every action, inside the boundary, under one control plane.
Retrieval honors the permissions of your source systems on every request. Nobody sees a document, clause, or matter they could not already open.
Every model call, tool invocation, and agent step is logged end to end, so each answer and action traces back to its source and its actor.
Critical actions pause for review. Firm playbooks decide what an agent may do on its own and what must wait for an attorney to approve.
Ingestion, indexing, and inference all run inside the enclave. Your knowledge is never used to train shared models and never crosses the boundary.
Reframe is one system in three deliberate layers. Security is the middle layer, the deployment that holds the other two inside your boundary.
Holds the knowledge. Your firm structured into a per-tenant Context Graph, grounded to the source and permission aware.
Keeps it yours. The single governed layer inside your tenant, your keys, your network, enforced at query time.
Puts it to work. Chat, Search, the Word plugin, and Agent Workflows, every answer grounded and every action audited.
The firms that cannot afford to guess have already stopped guessing.