252 Commits This Week
as of Nov 15, 2026

Open-Source Cloud and HPC Platform

One self-service marketplace for private clouds, public clouds, HPC and AI services.
Identity, billing, quotas and support desk are built in.

Overview

Service marketplace

Providers publish offerings with price plans. Customers order them, with optional approval by the project, organization or provider.

Self-service portal

One portal with separate workspaces for end users, service providers and staff. Users sign in with their institutional or national identity.

Reports and dashboards

Usage, cost, revenue and capacity reports for staff, providers and organization owners, all exportable. Metrics can also be sent to Prometheus and Grafana.

Accounting and billing

Fixed-price, usage-based, one-time and limit-based pricing, with automatic invoices or export to external billing systems.

Service desk

Built-in support tickets, or a connection to Jira Service Management, Zammad or OpenText SMAX, so customers have one support line for every service.

Open source

MIT-licensed. Deploy with Helm or Docker Compose, on premises or in the cloud, with professional support available.

See Waldur in action

Cloud, HPC and AI services from every provider in one catalogue, ordered in a few clicks with the price shown up front.

Trusted by leading organizations

ETAIS logo
GÉANT logo
EGI logo
EOSC logo
OCRE logo
University of Cambridge logo
University of Tartu logo
CESNET logo
CSC logo
KIFÜ logo
LUMI logo
NORDUnet logo

Features

One catalog for every backend

Offer private and public clouds, HPC clusters, Kubernetes platforms, storage and AI gateways side by side in one marketplace.
See all integrations

Teams, projects and roles

Customers invite people into projects, delegate resource management with role-based access and let roles expire automatically. Every change is recorded in an audit log.

Policies and budgets

Set cost and usage limits per organization, project or offering. When a limit is reached, Waldur notifies the team, blocks new orders or requests downscaling.
Learn more

Calls for proposals

Hand out resources through calls: application forms, review rounds, conflict-of-interest checks and allocation decisions. Accepted proposals become projects with the awarded resources.
Learn more

AI assistant

A built-in assistant helps users in plain language: finding open calls, preparing proposals, managing resources and finding their way around the portal.
Learn more

Project chat on Matrix

Every project gets an end-to-end encrypted Matrix room. Members join automatically based on their roles, and a Waldur bot answers questions about status, orders and members and opens support tickets from the chat.
Learn more

Automation and APIs

Everything in the portal is available through the REST API, with a Terraform provider, an Ansible collection and SDKs for Python and TypeScript.
Learn more

Credentials stay with the provider

Providers can connect their systems through the Waldur Site Agent, which runs on their side and keeps infrastructure credentials there.
Learn more

Federation

Independent Waldur deployments can share offerings, so partners use each other’s services while each keeps its own governance.
Learn more

Waldur HomePort Screenshots

From the catalog to the invoice, on a demo deployment with made-up organizations.
The catalog: an HPC partition, an OpenStack tenant, object storage and LLM inference from two providers
An HPC allocation on Slurm, with a year of usage reported by the site agent
An LLM inference resource and its API keys, each with a monthly token quota
One monthly invoice with tokens, core-hours, storage and GPU-hours from two providers
Revenue growth: what each organization paid per month over a year, and the total
Usage trends: this year's usage, its change from last year, and month by month
A SLURM usage policy: what happens at the limit, the period, carryover and the limit per component
The site agent of an HPC offering: its version, uptime and the services it runs

How it works

Waldur sits between the people who use cloud, HPC and AI services and the infrastructure that runs them: one portal and API in front, every provider's services behind it.

How Waldur connects people with the infrastructure behind its catalogPeople sign in through their own identity provider and work in the Waldur portal or its REST API. Inside Waldur are the catalog and orders, organizations and projects, accounting, quotas and policies, calls for proposals and the support desk. Scripts and AI agents use the API, and tickets sync with an external service desk. Clouds with an API, such as OpenStack, VMware vCenter, Rancher or another Waldur, are provisioned directly. Clusters and services such as Slurm, MOAB, Kubernetes and LiteLLM are run by the Waldur Site Agent at the provider's site: it fetches orders from Waldur, reports usage back and keeps the credentials on site.Usersorder and use servicesService providerspublish offeringsOperatorsrun the platformsign inYour identity providereduGAIN, Keycloak, LDAP, SAMLauthenticatesWaldurWeb portalREST APICatalog and ordersOrganizations,projects, rolesAccounting,invoicesQuotas andpoliciesCalls forproposalsSupport deskAutomationSDKs, CLI, Terraform,Ansible, AI agents (MCP)Service deskJira Service Management,Zammad, OpenText SMAXuses the APIsyncs ticketsClouds with an APIOpenStack, VMware vCenter,Rancher, another Waldurprovisionsthrough the APIAt the provider's siteWaldur Site Agentkeeps credentials on sitefetches orders,reports usagecreates accounts,sets limitsClusters and servicesSlurm, MOAB, Kubernetes, OpenNebula,Azure, DigitalOcean, Ceph, LiteLLM How Waldur connects people with the infrastructure behind its catalogPeople sign in through their own identity provider and work in the Waldur portal or its REST API. Inside Waldur are the catalog and orders, organizations and projects, accounting, quotas and policies, calls for proposals and the support desk. Scripts and AI agents use the API, and tickets sync with an external service desk. Clouds with an API, such as OpenStack, VMware vCenter, Rancher or another Waldur, are provisioned directly. Clusters and services such as Slurm, MOAB, Kubernetes and LiteLLM are run by the Waldur Site Agent at the provider's site: it fetches orders from Waldur, reports usage back and keeps the credentials on site.UsersServiceprovidersOperatorssign inYour identity providereduGAIN, Keycloak, LDAP, SAMLauthenticatesWaldurWeb portalREST APICatalog and ordersProjects and rolesAccounting, invoicesQuotas and policiesCalls for proposalsSupport deskAutomationSDKs, CLI, Terraform,Ansible, MCPService deskJira, Zammad,OpenText SMAXuses the APIsyncs ticketsprovisions through the APIClouds with an APIOpenStack, VMware vCenter, Rancher, WaldurAt the provider's siteWaldur Site Agentkeeps credentials on sitecreates accounts, sets limitsClusters and servicesSlurm, MOAB, Kubernetes, OpenNebula,Azure, DigitalOcean, Ceph, LiteLLMfetches orders, reports usage
Clouds with an API are provisioned directly. Clusters and services are run by the Waldur Site Agent at the provider's site, which only connects out to Waldur: no inbound ports, and the credentials stay on site.
Download as PDF This diagram and what each part does, on two A4 pages
  1. Providers connect their infrastructure

    OpenStack, VMware vCenter, Rancher and other Waldur deployments connect through their APIs. Slurm and MOAB clusters, Kubernetes, OpenNebula, storage and AI gateways connect through the Site Agent. Each provider then publishes offerings with their prices.

  2. Users order from one catalog

    People sign in with their institutional or national identity and order an offering for a project. An order can wait for approval by the project, the organization or the provider, or resources can be awarded through a call for proposals.

  3. Waldur runs it and keeps the books

    Waldur provisions the resource, keeps the project's members in sync on the backend and collects usage. Usage turns into monthly invoices, and policies warn the team, block new orders or request downscaling when a budget runs out.

Under the hood: a Django backend and a React portal, with PostgreSQL and RabbitMQ, deployed with Helm on Kubernetes or with Docker Compose on a single machine.

Integrations

Waldur works with the infrastructure you already run. Every backend plugs into the same marketplace, identity, accounting and support layer, so users get one catalog no matter what runs underneath.

  • OpenNebula
  • OpenStack
  • VMware vCenter
  • DigitalOcean
  • Azure
  • FirecREST
  • LEXIS
  • MOAB
  • Open OnDemand
  • OpenPortal
  • Slurm
  • Envoy AI Gateway
  • LiteLLM
  • Waldur MCP server
  • eduGAIN
  • Keycloak
  • LDAP
  • MyAccessID
  • SAML
  • Harbor
  • Kubernetes
  • OKD/OpenShift
  • Rancher
  • Jira Service Management
  • Matrix
  • OpenText SMAX
  • Zammad
  • Ceph RadosGW
  • croit
  • Lustre / CephFS
  • Nextcloud
  • Grafana
  • Prometheus
  • Ansible
  • SDKs
  • Terraform
  • waldur-cli
Browse all 58 integrations

Missing something? New integrations are built on demand, or with custom scripts.

Documentation and resources

Get started

Try Waldur on your own machine, deploy it for production with Docker Compose or Helm, and set it up for your users.
Get started

User guide

Step-by-step guides for every role: organization owners, service providers, call managers, end users and support staff.
Read the user guide

API and SDKs

Everything in the portal is available through the REST API, directly or through the SDKs for Python, TypeScript and Go, the command-line client, the Terraform provider and the Ansible collection.
API reference

Developer guide

How the backend and the portal are built, how to set up a development environment, and how to contribute.
Read the developer guide

Waldur escape room

Learn Waldur by playing: solve puzzles about its features to bring a cloud back after the Great Cloud Meltdown.
Play the game

Telemetry

Live statistics on Waldur deployments worldwide: how many there are, their users, countries and the versions in use.
Open the dashboard