LibraOS

The operating system for the AI workforce

Put company knowledge
to work with
digital employees.

Build and run AI agents with your company’s knowledge, permissions and review rules. Work with them in Libra Desk. Deploy LibraOS on your own infrastructure.

See how it works Book a demo

On-premise deployment · Cloud private beta
Build your own application →

A customer question. A response to review.Libra Desk
Desk interface excerpt showing a customer reply waiting for a reviewer’s decision
Customer contextPrepared replyHuman review
Actual Desk components · fictional demonstration data
Explore the workflow

Built for work across industries

Exabits AI infrastructureEqualDocs Legal technologyBona Sonority School Education

01 / Work with Libra Desk

Start with the work
in front of you.

Find a sourced answer, prepare a customer reply, or review a record update. Libra Desk brings agents into your team's everyday work, on the LibraOS runtime.

Turn company knowledge into useful answers

Find the answer.
Check the source.

A colleague asks what to prepare for a site visit. Use the relevant company guide, then inspect the passage behind the answer.

  1. Add the knowledgeAn admin uploads a document or adds text to organization knowledge.
  2. Give the agent accessChoose which knowledge the agent can use. Manage its access separately from the organization’s documents.
  3. Ask and verifyAsk your question, then expand a returned source to read the supporting passage.
See it with your knowledge
Libra Desk / KnowledgeInterface excerpt
Desk source viewer with the site preparation guide expanded beneath an example answer
Read the retrieved passage behind the answer.View larger

Example digital employee / Knowledge assistant

Responsibility
Answer questions about preparing for a site visit.
Knowledge
The assigned site preparation guide.
Allowed work
Retrieve relevant passages and prepare an answer.
Human judgment
The colleague checks the sources before using the answer.

Prepare a customer response

A helpful draft.
A human decision.

The same site-visit question arrives from a customer. Prepare a reply in context and review exactly what would be sent.

  1. Open the requestWork from the customer's message and conversation history.
  2. Prepare the replyUse the assistant’s draft and check the supporting source before approving a response.
  3. Review the actionAn authorized reviewer approves or rejects the proposed reply. Check the execution result to confirm the send action completed.
What happens after review?

Awaiting review: the reply is a proposal.

Sent: the send action completed.

Failed: inspect the error and prepare a new request if needed.

Unknown: refresh the status before taking another action.

Approval and execution have separate outcomes.

Walk through a support request
Libra Desk / Customer supportInterface excerpt
Desk proposed reply with its draft, named recipient, Approve and send button, and Reject control
Inspect the draft and recipient before deciding.View larger

Example digital employee / Customer support assistant

Responsibility
Prepare a response to a customer’s site-visit question.
Knowledge
The conversation and assigned company guidance.
Allowed work
Draft a reply for the named recipient.
Human judgment
An authorized reviewer decides whether to send it.

Keep customer records current

See the change
before it becomes fact.

A customer confirms interest in an on-site assessment. Review a proposed relationship update alongside the reason for it.

  1. Start with the recordReview the customer's details and the context of the relationship.
  2. Compare the proposalSee the current value, suggested value and explanation together.
  3. Decide what changesApprove, edit or decline. The service checks write access and records the decision.
Explore your sales workflow
Libra Desk / Sales operationsInterface excerpt
Desk CRM proposal comparing warm to hot relationship status, with rationale and Approve, Edit and Decline controls
Compare the proposed update with the current record.View larger

Example digital employee / Sales operations assistant

Responsibility
Propose an update to a customer record.
Knowledge
The current record and relevant conversation context.
Allowed work
Suggest a field change and explain the reason.
Human judgment
A reviewer with record access approves, edits or declines.

Desk interface excerpts with fictional demonstration data. These previews do not run agents or change records.

Explore the workspace
Libra Desk Where your team worksLibraOS The runtime underneathBuild your own application on the same foundation

02 / The platform

Give every agent
a foundation for work.

A digital employee is an AI agent with a defined responsibility, assigned knowledge and permitted tools. Your team sets the review rules. LibraOS provides the runtime that governs the work.

Know the business

Knowledge with context.

Connect company knowledge and persistent memory to the agents that need them. Grounded answers carry references to their sources.

Explore the runtime
Define the boundaries

Control how agents act.

Scope access and execution with permissions and policies. Route actions that need judgment to the people responsible for them.

Explore governance
Follow the work

A record you can inspect.

Keep the actor, requested action and execution outcome connected. Understand what happened and where human review was required.

Build on LibraOS

03 / Your deployment

Choose where
your agents run.

Match the deployment to your infrastructure, model choices and data requirements.

LibraOS Self-Hosted

Your infrastructure.
Your operating boundaries.

Deploy the runtime on your own infrastructure. With local models and connectors, configure an air-gapped environment.

  • Control the environment and model connections
  • Choose local models or approved external providers
  • Generate and review a starting team in setup
Explore self-hosted
LibraOS Cloud Private beta

A hosted path
to getting started.

Request access to the hosted edition. Join the private beta while we expand availability.

  • A managed deployment of the LibraOS runtime
  • Bring your team's work and governance requirements
  • Access coordinated with the LibraOS team
Join the Cloud beta
Compare deployment responsibilities and data boundaries

04 / In practice

One foundation.
Different businesses.

Customers build their own services and workflows on LibraOS. The application and domain expertise stay with the teams delivering them.

AI infrastructure

Exabits

A knowledge layer for a business operating AI data-center infrastructure.

Legal technology

EqualDocs

A customer building legal AI services with its own legal expertise and application.

Education

Bona Sonority School

Supervised agents supporting document work and education workflows.

Start with your work

Bring a workflow.
See where it can go.

Walk through your use case, knowledge sources and approval requirements with our engineering team.

Prefer to build? Start with the developer guide

Engineering contact only. Read our Privacy Policy.

Desk interface excerpt

Fictional demonstration data · image preview