Platform architecture
Four layers that every production enterprise AI solution needs: context, execution, governance and integration. Deployed privately, in your cloud or inside your client's environment.
Partner architecture
Domain workflows, proprietary logic, vertical data, UX and the customer proposition.
Context layer · agent execution · governance · evidence · integration · deployment.
Identity, systems of record, documents, APIs, data stores and the chosen models, in the cloud, on-premises or air-gapped.
You build the solution. The customer keeps its environment. SmoothOperator.ai runs the layer between them.
Trace a request through the layersTechnical request path
This is the logical flow of a partner solution on SmoothOperator.ai. It shows where context is scoped, when controls apply, and what can be reviewed; it is not a network-topology diagram.
The partner solution supplies the user experience and domain logic. The workflow begins with the user's request and its permitted operating context.
Collection, agent and workflow permissions scope which client knowledge is available. Client context remains separate from partner templates and other clients' collections.
Execution controls define the allowed actions while agents run, rather than relying only on a review of the final answer.
Orchestrated or specialised agents use the model access and authenticated, allowlisted tool connections agreed for the deployment.
Cited answers, agent steps, tool calls and structured outputs can be traced and exported for review against their sources.
Where execution runs, which systems are connected, and whether a model is external depend on the solution and the agreed deployment. This logical sequence does not assert that every request stays inside one network.
For workflows that use an external model, review the anonymization layer and the proposed data boundary during technical scoping.
For the actual hosting topology, model and tool configuration, access model, and evidence format, request a technical session for your proposed deployment.
Four production layers
01
Private context layer
A governed document and knowledge layer built from collections. Each collection sets what an agent can see and act on, by role. Terms like "customer", "policy" or "case" mean the same thing across every agent and workflow, without that knowledge ever sitting in a vendor's cloud.
Why it matters to partners
Each client's data stays separate, and your templates and rules stay separate from theirs.
02
Agent execution
Orchestrated agent teams for complex reasoning, and specialised agents for fast domain queries. Every run produces structured outputs (threads, attachments, exports) that your clients' teams can review, reuse and audit.
Why it matters to partners
You design the workflow once and deploy it to every client.
03
Governance and controls
RunControls define what each workflow may and may not do, and enforce it while the agent runs. Access is controlled at the collection, agent and workflow level. Every tool call, agent step and output is traceable.
Why it matters to partners
You can sell into regulated clients, where "the AI said so" is never an acceptable answer.
04
Open integration
Works with any model provider using your own keys. Native MCP tool server support, with consistent authentication and allowlists, lets agents act across your clients' existing systems. Switching models is a configuration change, not a migration.
Why it matters to partners
Your product never depends on one model vendor's roadmap or pricing.
Evidence-grade outputs
Answers are cited and checked for citation integrity. Any result exports as an audit-ready bundle that traces the question to the answer to the source.
Built for partners
| Capability | Status |
|---|---|
| Per-client deployment (cloud, on-prem, air-gap) | Available under partner agreement |
| Collection-level data separation per client | Available under partner agreement |
| Any model, bring your own keys | Available under partner agreement |
| Native MCP tool integration | Available under partner agreement |
| White-label and co-branded deployment | Available under partner agreement |
| Multi-tenant hosting with tenant isolation | Available under partner agreement |
| Per-tenant usage metering | Available under partner agreement |
| Partner administration console | Available under partner agreement |
| APIs and SDKs | Available under partner agreement |
Capability scope and deployment details are agreed for each partner solution.