Skip to content
Language: EN Español: ES
Let’s talk

Klarvai AI Governance Platform

All your AI, under control.

A workspace to get things done. A console to govern AI. An engine that turns each request into results, with your data and your rules.

Request a demo See how it works
  • Your infrastructure
  • Local or authorised models
  • Corporate identity

A space to work. A place to govern.

Control Center governs the AI: people, models, tools, data and budget. Workspace brings those decisions into daily work: chat, knowledge, files, projects and tasks.

Control Center on desktopWorkspace on mobile

Illustration with fictional organisations and data. The figures show demo activity, not customer results.

For your teamsWorkspace
Fewer tool switches. Chat, assistants, documents, files and tasks in one experience, with the models and plugins each person is authorised to use.
For IT and securityControl Center
Authorised models and tools, corporate identity, protected secrets and traceability. Each role sees only what it is allowed to see.
For leadership and financeReports and FinOps
Activity, consumption and budgets to decide where to extend AI and with what results.

Three decisions, one platform

  1. Where the AI runs.
  2. Which information it can use.
  3. Which tasks and actions it is allowed to perform.

Four capabilities. One platform.

A workspace for AI, with controls applied at the time of use. Workspace, Core, Runtime and execution services work under the same rules.

Access

One entry point for your team.

Workspace and corporate identity. People work with a familiar interface and their organisation’s permissions.

Control

Your policy in every request.

Approved models, limits and budgets. The gateway applies the rules before calling a model or tool.

Knowledge & Agents

AI connected to your business.

Knowledge with permissions and agents connected to your systems. Sensitive actions require human review.

Security & Assurance

Protection and traceability.

Input and output controls, usage metrics and evidence to review what happened.

  • People: a familiar interface.
  • IT: approved models and tools.
  • Security: data and actions under policy.
  • Leadership: usage, costs and results.

Klarvai Intelligence Engine

From a message to finished work.

The engine interprets the request, organises the work and uses authorised capabilities. Permissions, context and traceability stay with every step.

Klarvai Core authorises every action
  • Current permissions
  • Budget
  • Human review
01 / 04

Understands what you need.

The request becomes a specific task. The engine identifies the goal and prepares only the context the person is allowed to access.

Session and access are validated before work begins.
An example in Workspace

Prepare a report on this quarter’s outstanding invoices.

Request interpreted

01
  1. 01Goal: invoice report
  2. 02Context: Finance project
  3. 03Expected output: document
A clearly defined taskResult in Workspace
02 / 04

Maps the path to the result.

Organises steps and dependencies. Selects models and tools from the options your organisation permits.

The plan stays within the available permissions and limits.
An example in Workspace

Prepare a report on this quarter’s outstanding invoices.

Work plan prepared

02
  1. 01Read authorised invoices
  2. 02Group outstanding amounts
  3. 03Prepare the report with its sources
A plan you can followResult in Workspace
03 / 04

Acts with the right permissions.

Invokes models and tools, preserves progress and re-checks conditions before every action. Sensitive operations wait for approval.

An action awaiting approval cannot proceed on its own.
An example in Workspace

Prepare a report on this quarter’s outstanding invoices.

Execution under control

03
  1. 01Authorised system query
  2. 02Progress saved at checkpoints
  3. 03Approval before sharing
Traceable, recoverable workResult in Workspace
04 / 04

Checks and delivers.

Checks the registered file and its format. Compatible flows compare claims with sources; Workspace receives the result and the Core records consumption.

Checks provide evidence for human review.
An example in Workspace

Prepare a report on this quarter’s outstanding invoices.

Report available for review

04
  1. File generated and registered
  2. Format checked and sources available
  3. Activity and consumption attributed
A deliverable you can reviewFile, sources and activity in the same conversation.
WorkspaceKlarvai CoreRuntimeAI Gateway / Artifacts / Tools

Illustrative example. Checks help detect errors; reviewing the results remains your team’s responsibility.

Governance with evidence.

Each use of AI has an owner, a decision and a record. Policies are simulated before they are saved, Guard protects content before the model, and the audit log keeps who, when, what and the result, never the conversation itself.

Use-case registry
Purpose, owner, data, risk and evaluation for each case. Submit for review, approve, reject with a reason or withdraw. Whoever submits cannot approve their own case.
Policies and simulation
Local-only areas, admitted providers and regions, models per department, output token limits and monthly caps per person. See who is affected before saving.
Klarvai Guard
Detects identifiers, bank details, cards, emails, phones and known secret formats; blocks or pseudonymises before the chat model. Findings keep types and counts, never the values.
Audit and FinOps
Insert-only audit of who, when, action, object and result, exportable as CSV. Budget is reserved before each run and settled afterwards, by person, department, agent and model.

The registry provides traceability for AI governance. It is not a certification, an automatic legal classification or a complete compliance suite.

An architecture built around your organisation.

People work in Workspace; administrators govern from Control Center. Core authorises, Runtime executes and services connect models, tools and data within the agreed environment.

Your environment. Your rules.Private cloud / On-premise / Air-gap
Klarvai Workspace

Chat, agents, projects and files

Control Center

Models, policies, costs and audit

Identity, authorisation and context
Klarvai CoreThe authority behind every operation
  • Identity
  • Permissions
  • Licences
  • Budget
  • Secrets
  • Audit
Authorisation per actionmTLS
Klarvai Runtime

Governed execution

Intelligence Engine
  1. Classify
  2. Plan
  3. Execute
  4. Verify
Guard + KnowledgeContext, protection and recovery

Execution services

AI GatewayAuthorised models
ArtifactsDocuments without network access
Controlled ToolsTools with limits
Your organisation’s data
  • Database
  • Files and vectors
  • Events and secrets
Optional observabilitySecure infrastructureOnline / offline licensing
Permissions are validated before acting. Results and consumption are recorded.

  1. Core: decide

    Centralises identity, permissions, licences, budget and audit. Authorisation is checked before every action.

  2. Runtime: execute

    Intelligence Engine coordinates work with limits, checkpoints and recovery. Sensitive actions require approval.

  3. Services: connect and produce

    AI Gateway connects models; Artifacts generates files without network access; Controlled Tools applies quotas and external access controls.

  4. Data: persist and isolate

    Files, vectors, events and secrets remain under the organisation’s controls. Observability is optional.

Where your AI lives

Private cloud
Kubernetes in your account or agreed infrastructure, with reference Helm chart and Terraform.
Your own servers
Services and data inside your organisation’s environment.
Air-gapped
Signed offline package and local models, with licence renewal by file.
Mixed
Local processing and external providers chosen by policy: which data and tasks may go to each destination.

It connects with your office suite (email, calendar and files), Slack, Notion, GitHub, Jira and Confluence, Salesforce and HubSpot through read adapters, and with your ERP, CRM, databases and internal APIs through connectors and reviewed MCP servers. Each connection is configured with your permissions and validated for the agreed resources and operations.

Explore the platform components

Every component has a clear responsibility.

Each capability ships as part of an integrated suite operated under a single brand. Start with the essentials and add modules as usage grows.

Workspace

The workspace for people: chat, agents, projects, documents and tasks in one experience.

  • Chat and agents
  • Projects and files
  • Skills and tasks
  • Results and sources

Control Center

The console for administration, governance and oversight. Manage decisions about models, policies, people and costs here.

  • Models and policies
  • People and permissions
  • Evaluations and reports
  • Audit and FinOps

Core

The authority behind operations: validates identity, permissions, licence and budget before authorising execution.

  • Identity and permissions
  • Licences and FinOps
  • Vault and agent catalogue
  • Scheduler and audit

Runtime

Executes authorised work. Intelligence Engine classifies, plans, executes and verifies, preserving state and checkpoints.

  • Intelligence Engine
  • Limits and cancellation
  • Checkpoints and recovery
  • Authorisation per action

AI Gateway and connectors

Connects to local or authorised external models, and enterprise systems through reviewed MCP servers and APIs.

  • Model catalogue
  • Policy-based routing
  • Enterprise connectors
  • Credentials and permissions

Artifacts

Generates documents and files in an isolated service without network access, with registration and format checks.

  • Documents and presentations
  • Spreadsheets and PDF
  • Registered files
  • Reopening and verification

Controlled Tools

Tools with network controls, quotas and isolation. Access to public sources follows the environment’s policies.

  • Web and temporary files
  • Restricted Python
  • URL and network controls
  • Quotas and isolation

Guard and Knowledge

Document context with permissions, search, OCR and citations. Guard detects configured patterns before the chat model, blocking or pseudonymising according to policy.

  • Knowledge with permissions
  • RAG, OCR and sources
  • Protection before chat
  • Versioned policies

Data layer

Per-organisation persistence: database, files, vectors, events and secrets, with controlled access.

  • Organisation separation
  • Files and vectors
  • Events and checkpoints
  • Encrypted secrets

Licensing and infrastructure

Online or offline licensing, deployment in the client environment and optional observability for activity, consumption and errors.

  • Licence issuance and renewal
  • Containers and Kubernetes
  • Backup and recovery
  • Optional OpenTelemetry and Langfuse

Questions about the platform.

What exactly is the platform for?

So the whole company uses AI from one place and under the company’s rules. Employees work with chat, assistants and agents in Workspace; IT, security and management decide in Control Center which models, data and tools sit behind it, with what permissions and at what cost, and see everything logged.

Can our data stay in our infrastructure?

Yes, with local models and services configured. The platform deploys in a private cloud, on your own servers or air-gapped; the routes for embeddings, tools, search, observability and backups are reviewed as well. A web search needs internet access.

Which AI models can be used?

Local models or authorised providers, within the approved catalogue and the gateway’s compatibility. The company decides which are approved and for which groups. Klarvai Auto applies rules to choose among permitted models, with no implicit public-provider fallback for a local route; access, pricing, capacity and quality are reviewed when a model changes.

Does it replace the assistants we already use?

Not necessarily. Klarvai is the layer that connects people, models, data and systems under the company’s policies. It coexists with your office suite and lets you use several model providers through a single governed front door.

Can it act without approval?

Only where the policy and the tool type allow it. Sensitive actions require an approval tied to their arguments and re-authorisation at execution time. Initial native connectors are read-only; writes require an approved MCP server or an extension.

How is confidential information protected?

Documents keep their permissions when indexed, so the AI never uses what the person cannot open. Guard detects identifiers and secrets before the model and blocks or pseudonymises them, and the logging level of each application is agreed with the department: full, metadata only, anonymised or no content.

Does it guarantee compliance or correct answers?

No. It provides controls, records and evidence. Legal assessment, process acceptance and the review of results are carried out in the project.

How do we start and what does it cost?

With a pilot on a bounded group of people, cases and data; it can focus on chat and knowledge first. The subscription includes the software, users, modules and environments of the plan; implementation, support and optional modules have their own prices. Scope, prerequisites, timeline and price are provided in writing.

Klarvai AI Governance Platform

Let’s talk about AI in your organisation.

Tell us what you want to solve. We will help you define the next step, from a demo to a pilot with your team.

Contact us