01
Identity
Which employee, device, and agent initiated the request?
Private AI for enterprise
Control the AI layer.
Keep the systems that work.
vama is a privately deployed AI control and knowledge layer that works alongside the systems your organisation already uses.
vama
Enterprise ControlPolicy, routing, audit
vama ClawLocal execution
Existing systemsConnected, not replaced
The enterprise question
AI capability is arriving faster than AI governance.
Employees and agents need to work across email, documents, CRM, code, databases, and specialist systems. Organisations need to know what is happening at every step.
02
Data flow
What information is being retrieved, transformed, or sent outside the organisation?
03
Action
Which model, MCP, API, or tool was used—and what result did it produce?
vama does not ask you to replace every system.
It gives your organisation a controlled way for AI agents to work across the systems you already rely on.
It gives your organisation a controlled way for AI agents to work across the systems you already rely on.
Positioning
Keep the systems of record. Add a governed AI operating layer.
Without a control layer
AI access becomes fragmented
- Personal API keys and unapproved tools
- Unclear model and connector routes
- Inconsistent handling of sensitive information
- Limited reconstruction after an incident
With vama Enterprise Control
AI access becomes inspectable
- Approved models, MCPs, APIs, and connectors
- Central identity, policy, and credential handling
- Skills and SOPs applied consistently
- Governed requests, approvals, and outcomes recorded
Reference architecture
One AI control layer. Many systems underneath.
The customer’s existing stack stays in place. vama controls the routes used by enterprise Claw instances.
Employee devicevama Claw, local files, local policy checks
Mobile oversightDispatch, capture, approval, status
vama Enterprise
ControlIdentity · policy · Skills · credentials
MCP/API gateway · model routing
audit · visibility · approvals
ControlIdentity · policy · Skills · credentials
MCP/API gateway · model routing
audit · visibility · approvals
Existing systemsGoogle Workspace, Microsoft 365, CRM, ERP, databases
Approved endpointsLLMs, MCP servers, APIs, company services
Open agent access
Bring other AI agents to the same institutional knowledge.
vama Control can expose approved knowledge and actions through governed MCP or API interfaces—not only to vama Claw, but also to other AI agents the organisation chooses to use.
What Control can provide
Shared, permission-aware access
- Institutional knowledge retrieval through approved interfaces
- Read and write actions where the customer grants permission
- Connector, data-classification, approval, and audit policies
- Access for agents such as Tencent WorkBuddy, OpenAI Codex, OpenClaw, Hermes, or another approved agent
What remains different
Control is deepest with vama Claw
- VinTech controls the Claw software, its local execution model, and its enterprise connection to Control
- With third-party agents, Control can govern the exposed interface and actions, but not the agent’s internal reasoning, runtime, or endpoint environment
- That distinction should be explicit in the customer’s security and deployment design
One control plane, multiple agent choices.
Customers can preserve optionality while keeping institutional knowledge and permitted actions behind a governed vama Control boundary.
Customers can preserve optionality while keeping institutional knowledge and permitted actions behind a governed vama Control boundary.
What vama delivers
Connect. Govern. Observe. Execute.
01 · CONNECT
Bring existing systems
Use approved connectors and APIs instead of starting a disruptive migration.
02 · GOVERN
Apply policy
Control identity, data classification, model routes, Skills, and permissions.
03 · OBSERVE
See what happened
Record requests, decisions, approvals, tools, outcomes, and exceptions.
04 · EXECUTE
Let Claw do the work
Employees ask for outcomes; Claw performs the permitted workflow locally and through approved routes.
The trust model
Security is the environment AI works inside.
Deployment, connector configuration, model routing, and policy choices define the actual boundary.
Data sovereignty
Deploy privately and decide which data may leave the customer-controlled boundary.
Model sovereignty
Route local, private, or external models according to sensitivity, capability, cost, and policy.
Agent governance
Use sandboxed execution, approved Skills, controlled tools, human approvals, and audit trails.
Institutional knowledge
Make company knowledge usable by people and agents.
vama can work with existing sources, index selected content privately, or provide a focused private data foundation.
Retrieve from the systems already in use
Approved connectors let Claw retrieve relevant email, documents, records, tickets, code, and other business context while Control applies identity, permissions, and data-flow rules.
Index selected information inside the private deployment
Build a permission-aware knowledge layer for policies, procedures, meeting memory, customer context, technical documentation, and approved internal research.
Use a small private data-and-function substrate
The Private Workplace Pack focuses on core knowledge, structured records, documents, storage, bots, basic communication, and approvals—not every niche feature in a mature office suite.
The human experience
Ask Claw for an outcome. Keep the controls in the background.
Claw can retrieve approved context, apply the relevant company Skill, use the permitted model and connector, prepare the message in the required format, and pause for review when needed.
A Skill can define the required data sources, calculations, disclosures, output format, and sign-off path for a recurring report.
A user can provide a receipt and a natural-language instruction. Claw can extract the data, apply coding and evidence rules, create the claim, and route it for approval.
The working relationship
Over time, Claw becomes the way people work with the private layer.
As staff use Claw repeatedly, the system can build a richer, permission-aware understanding of their working context and recurring needs.
01
From tool use to delegation
Staff can ask Claw to read, create, update, calculate, route, and format information instead of manually operating every document, spreadsheet, or specialist screen.
02
Less switching, by choice
When a company adopts the Private Workplace Pack, Claw can increasingly use its private knowledge, documents, records, and basic functions on the employee’s behalf. Existing systems remain available where they are still the best fit.
03
Better with context
With appropriate permissions and transparency, aggregated workflow and behavioural signals can improve Skills, recommendations, routing, and organisational capability. This is governed personalisation—not unrestricted surveillance.
The direction is gradual, not forced.
People may gradually rely less on third-party office tools when Claw and the private Workplace Pack provide a more capable, private, and familiar way to get work done.
People may gradually rely less on third-party office tools when Claw and the private Workplace Pack provide a more capable, private, and familiar way to get work done.
Adoption
Start with the control problem. Expand only where it helps.
1
Control + Claw
Retain the existing office and business stack. Govern enterprise Claw, AI routes, Skills, and knowledge access.
2
Control + selected utilities
Add the minimal Workplace Pack for private knowledge, structured data, documents, storage, bots, or approvals.
3
Progressive expansion
Move selected functions into the private environment only when the customer sees a clear advantage.
Control the AI layer without replacing the systems that work.
vama is developed by VinTech for organisations that want capable, governed, and privately deployed AI.