September 2026
What's new
HTTP transport for the IONOS CLOUD MCP Server
September 11
The IONOS CLOUD MCP Server can now serve the MCP protocol over Streamable HTTP in addition to stdio. Start it with --transport http (or IONOS_MCP_TRANSPORT=http); the default remains stdio, required by desktop clients like Claude Desktop, Cursor, and VS Code. The HTTP endpoint does not encrypt traffic or check who is connecting. Keep it bound to loopback (127.0.0.1) so only your own machine can reach it, or, if remote clients need access, put it behind a reverse proxy that adds TLS and authentication.
Opt-in write operations in the IONOS CLOUD MCP Server
September 7
The IONOS CLOUD MCP Server can now create, update, and delete resources on Compute Engine and Managed Kubernetes. The server remains read-only by default: the 77 write tools register only when you set the IONOS_MCP_TOOL_SCOPE environment variable to write or destructive, and every create, delete, and disruptive action takes a two-step confirmation.
Write operations create real, billable resources, and deletions cannot be undone. You are responsible for everything created, modified, or deleted in your account through these tools, and for the resulting charges.
User Privileges for Nextcloud Workspace Management
September 4
You can now control which IONOS CLOUD users in your organization can provision, edit, and delete Nextcloud Workspaces. A new DCD group privilege called Access and manage Nextcloud Workspace allows you to delegate infrastructure-level administration. This privilege governs administrative access at the IONOS CLOUD level only, and it does not affect user, group, or permission management within your Nextcloud Workspace instance.
DBaaS MariaDB v2 is now available in the DCD
September 3
You can now create and manage IONOS CLOUD DBaaS MariaDB v2 clusters in the Data Center Designer (DCD). MariaDB v2 supports MariaDB Long-Term Support (LTS) versions, including 10.6, 10.11, 11.4, 11.8, and 12.3. The DCD supports the full lifecycle of v2 clusters, including cluster creation, configuration updates, backup-based restore, and cluster deletion.
New Block Storage performance classes
September 1
Block Storage offers three new performance classes for Virtual Data Centers (VDCs):
Essential: Cost-efficient, for infrequently-accessed data and backups.
Balanced: General-purpose, balancing cost and performance.
Performance: High-throughput, for databases and latency-sensitive applications.
Existing VDCs storage types are unaffected. The following changes apply to new and existing VDCs:
New storage created through the DCD supports only the new Storage Performance classes.
These new performance classes rename the previous HDD and SSD options; the underlying technology is unchanged.
NIC Multi-Queue for Network File Storage
September 1
IONOS CLOUD Network File Storage (NFS) clusters now automatically use NIC Multi-Queue, distributing network traffic across multiple queues based on available CPU cores to improve throughput and reduce latency. The feature is activated by default on NFS clusters with at least two CPU threads (two cores, or one core with hyperthreading) and requires no additional configuration.
Documentation updates
Nextcloud Workspace Documentation
September 11
The Nextcloud Workspace documentation now covers four new topics. Applications lists the applications that are active from the start and those a workspace administrator can enable. Nextcloud Assistant describes the AI capabilities, how they are billed, and how data is handled. Backup and Recovery sets out the difference between workspace restore, file versioning, and backup. Security and Compliance covers hosting, data location, encryption, access management, updates, and certifications.
The FAQ has also been expanded from 7 to 32 answers, adding sections on domains, migration, storage, billing, and access management.
Image Input Guide for AI Model Hub
September 10
A new how-to guide brings image input for AI Model Hub together in one place. It lists every endpoint that accepts an image, shows how to attach one as a remote URL or as an inline base64 data URI, and sets out the accepted formats, the request size limits, and how images affect token cost.
MariaDB cluster metrics
September 8
New documentation covers infrastructure and database health metrics for MariaDB v2 clusters. You can view them in the DCD by following View MariaDB Cluster Metrics, or through the API, as described in Monitor MariaDB v2 Clusters Using Grafana.
Compliance Documentation for AI Model Hub
September 4
The two compliance pages for AI Model Hub have been revised. The EU AI Act page now sets out how the roles under the Act are allocated between the model developers as providers of the AI models, IONOS CLOUD as provider of the AI system, and you as deployer. It also adds a notice on the Article 50(2) marking obligation for AI-generated content, and a subsection allocating each of the Article 50 obligations.
The Data Handling page has been rewritten and now describes in more detail what happens to your data when you use the service.
Cost and Usage FAQs for AI Model Hub
September 4
The AI Model Hub FAQ now explains how usage is measured for each type of model and how to retrieve your usage through the IONOS CLOUD Billing API.
Query Monitoring Service Metrics from Your External Grafana Instance
September 1
A new tutorial demonstrates how to connect your external Grafana instance to the IONOS CLOUD Monitoring Service using the internal Grafana data source proxy. The tutorial covers creating a service account token in IONOS CLOUD Grafana, locating the Grafana Mimir datasource Unique Identifier (UID), and configuring the external Grafana instance with the correct authentication and datasource URL.
Responses API for AI Model Hub
September 1
A new how-to guide covers the AI Model Hub Responses API at /v1/responses, an alternative to chat completions that takes a flat list of input items and returns typed output items. It shows how to send a request, read the typed output (including the model's reasoning trace), call a tool and return its result, and lists which parts of the OpenAI Responses API are not available.
Last updated
Was this helpful?