Skip to content

CHANGELOG.md

Latest commit

 

History

History
74 lines (46 loc) · 8.5 KB

CHANGELOG.md

File metadata and controls

74 lines (46 loc) · 8.5 KB
Raw

Changelog

Unreleased

Added

  • Write operations for DNS, off by default and behind the same IONOS_MCP_TOOL_SCOPE gate: create, update and delete for primary zones, records, secondary zones and reverse records, plus DNSSEC enable/disable, start_dns_zone_transfer and import_dns_zone_file. Every create and delete takes two calls — the first returns a preview and a one-time token, the second carries it. The 14 existing DNS read tools now carry the readOnlyHint annotation.

    import_dns_zone_file replaces every record in a zone, so it requires destructive rather than write despite being a PUT, and its preview counts the records it would replace. delete_dns_zone_dnssec_key warns that the registrar's DS record must be removed first — a stale DS record makes validating resolvers answer SERVFAIL for the whole zone. Updates are PUT, so each one reads the resource and carries every field it does not change forward, including the identity fields (a zone's name, a record's name and type, a reverse record's IP), which are therefore not tool inputs.

1.1.0

Added

  • Write operations for Compute Engine, off by default. Opt in with IONOS_MCP_TOOL_SCOPE: write enables create and update, destructive also enables delete. Covers data centers, servers, volumes, NICs, LANs, firewall rules, security groups and their rules, IP blocks, classic/network/application load balancers and their forwarding rules, NAT gateways, target groups, cross connects, snapshots and images — plus server power control (start, stop, reboot, suspend, resume, upgrade), volume snapshot create and restore, and volume attach/detach.

    Creating and deleting takes two calls: the first returns a preview of exactly what will change plus a one-time token, the second carries that token and executes. A delete preview also lists what else it affects — the servers and volumes inside a data center, the NICs on a LAN, the backends behind a forwarding rule.

    ⚠️ These tools create real resources that cost money, and deletions cannot be undone. Read the warning in the README before enabling them. Enabling them is done at your own risk; IONOS Cloud is not responsible for actions the model takes as a result.

  • Write operations for Managed Kubernetes, off by default and behind the same IONOS_MCP_TOOL_SCOPE gate: create, update and delete for clusters and node pools, plus recreate_k8s_node and delete_k8s_node. Node pool creation covers autoscaling, attached LANs with static routes, labels, annotations and reserved public IPs. Every create and delete takes two calls — the first returns a preview and a one-time token, the second carries it. All of these are asynchronous, so poll get_k8s_cluster or get_k8s_nodepool and read metadata.state before chaining a dependent call.

    Three IONOS API limits to be aware of: a node pool cannot be renamed, an autoscaler's bounds can be changed but it cannot be switched off, and delete_k8s_node is not a reliable way to replace a node — it removes the node first, and an active autoscaler may hold the pool at the smaller size, so prefer recreate_k8s_node.

  • Read-only annotations on the Kubernetes read tools: the 8 existing list_*/get_* Kubernetes tools now carry the readOnlyHint annotation that clients use to decide whether to prompt before a call. No behaviour or signature change for callers.

  • filters parameter on all compute list_* tools: every compute list tool now accepts an optional filters object ({"property": "value"} pairs) that is forwarded to the API as server-side query params, so only matching resources are returned. Useful for scoping large result sets by name, location, state, image type, and more without client-side post-processing. If a filter returns an empty result, retry without it — a typo or value mismatch silently returns nothing. Filterable properties vary by resource (e.g. name, location, vmState, cpuFamily, imageType).

v1.0.1 — June 2026

Added

  • optional depth parameter on compute and Kubernetes tools: every list_* tool now defaults to depth=1, returning resource names and properties in a single API call instead of only UUIDs. Every get_* tool accepts an optional depth to control nesting. Eliminates the agent loop of N follow-up get_* calls when resolving names.

  • dynamic load mode (alias search): exposes only three meta-tools — ionos_search_tools, ionos_describe_tools, ionos_call_tool — through which the model discovers and invokes the full catalogue at runtime. The public tool list never changes (no notifications/tools/list_changed needed), so it fits clients with hard tool caps and no tool search of their own (Cursor ~40, Windsurf 100). Select with --load-mode dynamic or IONOS_MCP_LOAD_MODE=dynamic.

  • --load-mode flag: selects the tool-registration strategy (eager | lazy | dynamic) from the command line. Takes precedence over IONOS_MCP_LOAD_MODE; the effective mode and its source are logged to stderr at startup.

  • Managed Kubernetes: 8 new read-only tools as a dedicated product (eagerly loaded) — list_k8s_clusters, get_k8s_cluster, list_k8s_nodepools, get_k8s_nodepool, list_k8s_nodepool_nodes, get_k8s_node, list_k8s_versions, get_k8s_default_version. Covers the full cluster/node-pool/node hierarchy plus available version discovery.

v1.0.0 — June 2026

First release. An MCP server that lets LLM clients explore an IONOS CLOUD account read-only.

Added

  • Homebrew install: brew install ionos-cloud/ionos-cloud/ionoscloud-mcp (macOS + Linux, Intel + Apple Silicon).
  • Docker image: ghcr.io/ionos-cloud/ionoscloud-mcp (multi-arch: linux/amd64, linux/arm64). Pull and run with docker run -i --rm -e IONOS_TOKEN=… ghcr.io/ionos-cloud/ionoscloud-mcp.
  • Pre-built binaries on every GitHub release for linux, macOS, and Windows (amd64 + arm64).
  • --version flag: prints the server version. Useful for bug reports and version pinning.
  • Compute Engine: Browse datacenters, servers, volumes, NICs, LANs, IP blocks, load balancers (ALB/NLB), NAT gateways, security groups, firewall rules, target groups, private cross-connects, snapshots, images, templates, locations, server GPUs and CD-ROMs, remote console URLs, and provisioning request status.
  • DNS: Inspect zones, zone files, records, reverse records, secondary zones (incl. AXFR transfer status), DNSSEC keys, and account quota.
  • Billing: View your contract and billing profile, invoices, product catalog, traffic and usage breakdowns (per datacenter, per period), and daily utilization.
  • Object Storage: List buckets and objects across regions, read every bucket configuration surface (CORS, encryption, lifecycle, locking, policy, public access, replication, tagging, versioning), object metadata, retention and legal hold state, access keys, and regions.
  • Certificate Manager: List certificates and auto-certificates, look up issuance providers.
  • Activity Log: Look up accessible contracts, query the audit trail of API requests (who did what, when, on which resource). Defaults to the last 7 days and excludes noisy async-provisioning echoes — both overridable.

Try

  • Cost audit: "Audit my IONOS CLOUD account, find the top 5 cost-inducing resources this month, and suggest cost-efficiency tips."
  • Security sweep: "List every bucket whose public access block is off or whose policy is public — flag anything that looks unintentional."
  • Audit trail: "Show me every failed API request on my contract in the last 30 days, grouped by user."
  • Forgotten resources: "Find unattached volumes, unused IP blocks, and stopped servers across all my datacenters."
  • DNS sanity check: "List all zones on my account and flag any without DNSSEC enabled or with records pointing to IPs I no longer own."
  • Certificate expiry: "Which certificates on my account expire in the next 60 days?"
  • Traffic spike investigation: "My last invoice was higher than usual — show me daily traffic and utilization for the previous billing period and tell me what changed."
  • Onboarding tour: "Walk me through what I have running on IONOS CLOUD — datacenters, servers, storage, DNS — like you're explaining it to a new teammate."

Notes

  • Auth via the IONOS_TOKEN environment variable. Server refuses to start without it.
  • All responses are compacted before being returned to the LLM (no pretty-print, redundant audit envelopes stripped) so a single tool call uses far fewer tokens.
  • Brand: tool descriptions and docs consistently use "IONOS CLOUD".