Platform
Useful AI needs context. Trusted AI needs control.
The context plane tells your AI what your business knows. The action plane connects it to the systems where work gets done. The control plane surrounds both and decides who can see what, which actions are allowed, and what gets logged. Each plane is tailored to your organization and deployed in your own cloud.
Context Plane
What AI knows.
The right knowledge, assembled for each request.
The Context Plane turns your documents, records, and systems into structured, connected knowledge, then assembles a complete context package for every request, limited to what the requester is allowed to see.
Graph-augmented RAG
Retrieval that follows relationships between entities, not just text similarity, so answers draw on connected facts rather than isolated passages.
Knowledge graph
Customers, products, policies, accounts, and processes modeled as entities and relationships drawn from your systems of record.
Memory
Durable, scoped memory for users, teams, and workflows, governed by the same permissions and retention rules as the rest of your data.
Ontology
A shared vocabulary for your business: the terms, definitions, and rules that keep every assistant consistent with how your organization talks and works.
Context packages over MCP and API
Complete, cited context bundles delivered through the Model Context Protocol and a REST API, ready for any assistant, agent, or model.
Action Plane
What AI can do.
Useful work, through interfaces you already control.
The Action Plane exposes your systems to AI as well-defined, governed capabilities, so assistants and agents can look things up, draft, file, and update, without bespoke integrations for every use case.
MCP
Your capabilities published as Model Context Protocol servers, so compatible assistants and agent frameworks can discover and use them consistently.
APIs
Typed, versioned APIs wrapping internal systems, with inputs and outputs validated before anything reaches a system of record.
Tools
Purpose-built tools for search, lookup, drafting, and transactions, each with a clear contract and a clear owner.
Agents
Multi-step agents for well-scoped workflows, composed from approved tools and subject to the Control Plane at every step.
Control Plane
What AI is allowed to know and do.
Governance that is enforced, not just documented.
The Control Plane wraps both other planes, so nothing reaches the Context or Action Plane without passing through it. It decides, for every request and every action, who is asking, what they may see, what they may do, which model may handle it, and what gets recorded.
IAM and SSO
Integrates with your identity provider so every request carries a real, verified identity.
Role-based permissions
Entitlements mirrored from source systems and enforced at retrieval time, not after the fact.
Policy
Declarative rules for what data, tools, and models each role and use case may access.
DLP
Detection and redaction of sensitive data in prompts, context, and responses.
Audit and lineage
A traceable record of which sources, tools, and models produced each answer or action.
Approvals
Human-in-the-loop checkpoints for actions that carry risk or require sign-off.
Model routing
Send each request to an approved model based on sensitivity, task, and cost.
Cost controls
Budgets, quotas, and usage visibility by team, application, and use case.
Guardrails
Input and output checks that keep assistants on-topic, on-policy, and within scope.
How it fits together
Every request passes through all three planes.
- 01
Request
A user or agent asks a question or requests an action, carrying a verified identity.
- 02
Authorize
The Control Plane resolves roles, entitlements, and policy for this request.
- 03
Assemble
The Context Plane retrieves only permitted knowledge and builds a cited context package.
- 04
Act
The Action Plane runs approved tools, pausing for human approval where policy requires it.
- 05
Record
Sources, tools, models, and decisions are logged with full lineage for audit.
Deployment
Runs in your cloud, inside your VPC.
Contexor is deployed into your own cloud environment. Your documents, indexes, knowledge graph, and logs stay inside your network boundary, under your security controls.
- Deployed in your cloud account and VPC, not a shared multi-tenant service
- Connects to your identity provider for SSO and role mapping
- Use the model providers and endpoints your organization has approved
- Logs and audit records written to storage you own and retain
- Infrastructure defined as code, reviewable by your platform and security teams
Evaluation
Evaluated against plain RAG.
A context engine has to earn its place. In every engagement we build an evaluation set from your own questions and documents, and compare Contexor against a plain retrieval-augmented generation baseline on the same models and data.
The results are yours. They tell you where the engine helps, where it doesn't yet, and they become regression tests that run whenever content, models, or configuration change.
Answer quality
Correctness and completeness against reviewed reference answers.
Grounding
Whether each claim is supported by the cited sources.
Permission safety
Whether any answer draws on content the user should not see.
Latency and cost
Response time and spend per request, compared side by side.
See the three planes on your own data.
Bring one high-value use case. We scope it, build it in your environment, and measure it against plain RAG, so you can decide on evidence.