Three ways to work with the platform
Chat
A conversation with the model: a question, a document, an image. The fastest route to value and usually the first step of a deployment.
Agentic work
Multi-step tasks: the agent plans, reaches for data, checks the result and hands over a finished artefact for approval.
Automation
Workflows triggered by a schedule or an event, with an approval gate before any write operation.
Modules
Qosmo ChatChat, agentic work and automation; images and PDFs, model choice per conversation
Qosmo FlowsVisual no-code builder for agents and automation
Qosmo TracerFull trace of prompts and model decisions; evaluation, costs
Qosmo DocsRepository for documents and search data, highly available
Qosmo AuthOne identity: OIDC/SAML, MFA, RBAC, AD and LDAP federation
Models
Six variants prepared for different classes of work, all served locally and selected per conversation or per workflow. Alongside them, the same platform can run open-weight models.
QosmoGeneral purpose model
Qosmo CoderCode, scripts, automation
Qosmo SecurityAudits, policies, compliance analysis
Qosmo OCRScans, forms, PDF files
Qosmo MedSoonSupport for clinical staff
Two deployment options
| On-premise | Cluster operated by QAI | |
|---|---|---|
| Where it runs | In your server room, on your hardware | On infrastructure provided by QAI, dedicated to you |
| Air-gapped | Fully possible | Depending on the agreed scope |
| Hardware | On your side, sized with a partner | On our side |
| Who it suits | Hospitals, essential entities, large organisations | Smaller units without a server room of their own |
Security and compliance
- a dedicated, isolated cluster for every client, with no sharing between organisations;
- one identity across the platform, federated with your user directory and multi-factor authentication;
- an audit trail of every query and model decision, required when documenting AI use;
- retention and data location under your control;
- encryption at rest and in transit, network policies that deny connections by default.