Admins and people
in the same portal.
Admins and users interact with the platform from the portal or with venticli: who gets in, who uses which model, how much credit each one has — and how you resell it under your brand if you act as a provider.
Who uses what,
is up to you.
The portal runs on your cluster. From here you invite people, assign quotas, choose models, approve remote agents and see who did what.
People, groups and quotas
Members with role and available quota, groups with reserved budget, and invitations. Sign in with email and password, Google or a passkey; a TOTP or passkey second factor is optional.
Credit and allocations
Total balance, quotas reserved to groups and what remains unallocated. Automatic recurring allocations; every movement has its place in the ledger.
Catalogue and Switchyard
The catalogue shows only what you are authorised to use: platform models, aliases and your own keys. Switchyard picks among authorised candidates by suitability, effort and cost; PII casting is active when policy requires it.
Remote agents with approval
A runtime on your hardware appears awaiting approval: you compare the SHA-256 fingerprint read on the machine and approve models and capabilities. An offline agent leaves the catalogue.
Certified runtimes
API compatibility is not a certification: only versions and functions that were tested are marked as such. In the current release, Strata and llama.cpp; vLLM, Ollama and SGLang are not certified.
Credit, keys,
API lab.
Everyone sees their own credit, creates keys with spending limits and tries every gateway operation in the Lab, under the same policies as production. Open WebUI chat uses the same credit and quotas.
Workspace overview
Available and reserved credit, usage over the last seven days by category, pending decisions and most-used models.
API lab
Try Chat, Responses, Messages, embeddings, speech, transcription, FIM, forecasting and files with your session. It spends real credit and generates the request in curl, Python or TypeScript.
Keys with limits
Each key has its allowed models, a per-request maximum, a total budget and an expiry. The secret is shown only at creation; you can disable, regenerate or revoke it.
Credit and usage
Top-ups by Stripe or bank transfer: credit becomes available only once payment is confirmed. Reservations, usage and workspace credits with exact amounts, VAT excluded.
Who gets in, and with which key
The people, groups and keys of each company, one selected workspace at a time.
People
Role and available quota for each person, with search and a row menu.
Groups and quotas
A group receives reserved credit and distributes it; recurring allocations can be scheduled.
Invitations
Email invitation with an initial role; the email queue is handled by the system.
API keys
Allowed models, per-request maximum, budget, expiry; instant regeneration and permanent revocation.
Sign-in
Email and password, Google or a passkey, with an optional TOTP or passkey second factor, through Keycloak OIDC.
Company review
Company identity and evidence are reviewed by an operator before payments and reseller features unlock.
Who can do what, decided by the server
The menu helps you find features, but every operation is authorised by the server: a direct link grants no privileges.
Four customer roles
Member, group manager, company administrator and reseller administrator, with distinct scopes.
Platform roles
Separate from company roles: being admin of your own company does not make you a platform admin.
Cascading policy
Platform, company, groups and user: higher constraints also apply to more specific configurations.
Permission check
Ask the service whether your session can perform an action, with an explanation of the decision.
What runs, for whom, at what price
Cloud, local runtimes and managed models in the same catalogue, under the same rules.
Catalogue and mappings
Public names kept apart from the upstream target; each feature depends on what the target supports.
Upstream providers and own keys
Encrypted personal credentials, explicit sharing and governed service fees.
Effective policies
Allowed models, default model, PII required and per-request maximum, per scope.
Switchyard
Selection among authorised candidates by suitability, effort and cost.
Managed inference
Compute pools, deployments, routing and autoscaling with approved plans and limits.
Price list and commissions
Public prices kept apart from verified tariffs and FX; reseller commissions from settled usage.
How much is used, and who changed what
Not conversation content: operational telemetry and a record of administrative actions.
Requests and performance
Requests, errors and latency for the workspace, with a link into Perses.
Traffic and usage
How much is used per model and period, computed from the available events.
Logs and traces
Operational logs and slow OpenTelemetry traces, filtered by company.
Activity log
Append-only: who acted, on which resource and which fields changed.
Your hardware and automation
Local runtimes and declarative resources are part of the portal, not separate tools.
Remote agents
Invitation, pairing, fingerprint approval and model publication, with explicit consent to sharing.
Certified runtimes
An honest list of what has been tested and what is merely compatible.
venticli
Validate, diff, apply and wait on declarative resources, with the same authorisations as the portal.
OpenCode
The launcher reads the authorised catalogue and refreshes it at startup.
Your brand,
our platform.
You administer your own company, or run an installation with your own customers, models and prices.
One installation, many tenants: capacity and costs depend on your infrastructure and the upstream AI services.
Application multi-tenancy
Each customer is a tenant: data, people, quotas and catalogue are separate, without a cluster per customer.
Resellers and commissions
Administration delegation and commissions calculated on settled usage.
Provider profile
Branding, domains, contacts and legal documents supplied by the owner. Open WebUI keeps its own branding and licence.
Payments and invoices
Top-ups by Stripe or bank transfer, invoices and per-tenant usage.
Get in touch
to get started.
We help you install Ventic on your infrastructure and start directly from the workload you want to put in production.