Logo

Command Palette

Search for a command to run...

MCP tool reference

The complete list of the 33 tools an AI agent can call on the Cloud 66 MCP server.

How to read this page

Each tool lists the arguments it takes and the scope its token needs. Tools your token isn't scoped for don't appear in your agent's tool list at all — see Permissions and scopes.

A few conventions apply throughout:

  • Every tool except list_accounts requires account_id. Call list_accounts first to get one.
  • IDs are UIDs, not database numbers. Get them from the corresponding list_ tool.
  • "Applications" are what the API and dashboard also call stacks.
  • List tools are paginated. They all accept limit (default 25, maximum 100), starting_after, ending_before and sort_direction (asc or desc). Cursors come from the previous response's next_cursor / previous_cursor.
  • Approval-gated tools pause until you confirm them in your browser. See Approving sensitive actions.

Accounts and identity

list_accounts

Lists every Cloud 66 account you belong to. This is the starting point for everything else.

  • Arguments: pagination only
  • Scope: account:list

get_account

Returns the details of one account.

  • Arguments: account_id
  • Scope: account:get

get_me

Returns your own user profile.

  • Arguments: account_id
  • Scope: me:list

Applications

list_applications

Lists the applications in an account, optionally filtered.

  • Arguments: account_id; optional environment (for example production), name, pagination
  • Scope: application:list

get_application

Returns the details of one application.

  • Arguments: account_id, application_id
  • Scope: application:get

delete_application

Permanently deletes an application and all its resources. This cannot be undone.

  • Arguments: account_id, application_id
  • Scope: application:delete
  • Requires browser approval with two-factor authentication

Deployments

list_deployments

Lists recent deployments for an application, with their status.

  • Arguments: account_id, application_id; optional pagination
  • Scope: deployment:list

get_deployment

Returns the details of one deployment.

  • Arguments: account_id, deployment_id
  • Scope: deployment:get

get_deployment_logs

Returns log entries for a deployment, with timestamps and severity levels.

  • Arguments: account_id, deployment_id; optional include_children (default true), pagination
  • Scope: deployment:get

create_deployment

Deploys an application. Either name a deployment profile, or pass individual deployment parameters — not both.

  • Arguments: account_id, application_id; optional deployment_profile (mutually exclusive with the parameters below), git_ref (branch, tag or commit SHA — not for container applications), services (container applications only, as name or name:tag), deploy_strategy, rollout_strategy, canary_percentage (0–100, canary rollouts only)
  • Scope: deployment:post
  • Asynchronous. Returns an operation handle; poll it with get_operation.

Deployment profiles

list_deployment_profiles

Lists the deployment profiles defined on an application.

  • Arguments: account_id, application_id; optional pagination
  • Scope: deployment_profile:list

get_deployment_profile

Returns the details of one deployment profile.

  • Arguments: account_id, deployment_profile_id
  • Scope: deployment_profile:get

create_deployment_profile

Creates a deployment profile on an application.

  • Arguments: account_id, application (the application UID), name (unique within the application); optional deploy_strategy, rollout_strategy, canary_percentage, git_reference, builds and publishes (container applications only), and the deploy toggles apply_security_upgrades, apply_infrastructure_upgrades, force_infrastructure_upgrades, auto_update_docker_versions, pause_and_unpause_jobs, run_deploy_command, reboot_servers, take_server_snapshots
  • Scope: deployment_profile:post

update_deployment_profile

Updates an existing deployment profile. The default profile can't be modified.

  • Arguments: account_id, deployment_profile_id; the same optional fields as create_deployment_profile, plus name to rename it (canary_percentage here is 1–99)
  • Scope: deployment_profile:put

delete_deployment_profile

Deletes a deployment profile. The default profile can't be deleted.

  • Arguments: account_id, deployment_profile_id
  • Scope: deployment_profile:delete
  • Requires browser approval with two-factor authentication

Servers

list_servers

Lists the servers belonging to an application.

  • Arguments: account_id, application_id; optional pagination
  • Scope: server:list

get_server

Returns the details of one server.

  • Arguments: account_id, server_id
  • Scope: server:get

list_server_groups

Lists an application's server groups — the roles servers are organised by, such as web, database or process.

  • Arguments: account_id, application_id; optional pagination
  • Scope: server_group:list

get_server_group

Returns the details of one server group.

  • Arguments: account_id, server_group_id
  • Scope: server_group:get

Clouds

list_clouds

Lists the cloud provider credentials registered on an account.

  • Arguments: account_id; optional pagination
  • Scope: cloud:list

get_cloud

Returns the details of one cloud.

  • Arguments: account_id, cloud_id
  • Scope: cloud:get

list_cloud_regions

Lists the regions available for a cloud, from the local catalogue — no call is made to the provider. Use it to pick a region for create_csv3_cluster.

  • Arguments: account_id, cloud_id
  • Scope: cloud:get

list_cloud_sizes

Lists the server sizes available for a cloud in a given region, from the local catalogue. Use it to pick sizes for create_csv3_cluster server templates.

  • Arguments: account_id, cloud_id, region
  • Scope: cloud:get

create_cloud

Registers a cloud provider account. The agent supplies only a name and a provider type — you enter the credentials in your browser.

  • Arguments: account_id, name (3–60 characters, unique per provider within the account), type (one of aws, digitalocean, googlecloud, azure_rm, linode, vultr, hetzner)
  • Scope: cloud:post
  • Requires the browser credential handoff. Asynchronous — returns an operation handle to poll with get_operation; the new cloud's ID appears when the operation succeeds.

DNS providers

list_dns_providers

Lists the DNS providers configured on an account.

  • Arguments: account_id; optional pagination
  • Scope: dns_provider:list

get_dns_provider

Returns the details of one DNS provider. Requires account ownership.

  • Arguments: account_id, dns_provider_id
  • Scope: dns_provider:get

Clusters

list_clusters

Lists the Kubernetes clusters in an account.

  • Arguments: account_id; optional pagination
  • Scope: cluster:list

get_cluster

Returns the details of one cluster.

  • Arguments: account_id, cluster_id
  • Scope: cluster:get

list_server_pools

Lists a cluster's server pools (its node pools).

  • Arguments: account_id, cluster_id; optional pagination
  • Scope: server_pool:list

list_cluster_databases

Lists the managed databases running on a cluster.

  • Arguments: account_id, cluster_id; optional pagination
  • Scope: cluster_database:list

create_csv3_cluster

Creates a new k3s cluster on an existing cloud.

  • Arguments: account_id, name (3–60 characters, unique within the account for the environment), cloud_id (a DigitalOcean, Vultr or Hetzner cloud), region (a provider region ID from list_cloud_regions), availability, server_templates, servers; optional environment (default production) and cluster-level vendor.config.network_identifier to place the cluster in an existing private network (DigitalOcean and Hetzner only)
  • Scope: cluster:post
  • Asynchronous. Returns an operation handle; poll it with get_operation.

Node pools are described by two linked arrays:

  • server_templates[] — named hardware specs, each { name, vendor: { type, config: { size } } }. Sizes come from list_cloud_sizes for that cloud and region.
  • servers[] — the pools themselves, each { role, template, count, name? }, where template names a server_templates[] entry.

Every template must be used by at least one pool, and every pool must name a real template. Topology depends on availability:

availabilityRules
reduced_availabilityExactly one worker pool, no manager pool, single control-plane node
high_availability (recommended)One or more worker pools, plus at most one manager pool with a count of 3, 5 or 7 and no name; worker sizes must be at least 2 vCPU / 2 GB

Total worker nodes across all worker pools must be between 1 and 15. Worker pool names are lowercase letters, digits and hyphens, and must be unique; manager pools must not have a name.

create_csv3_application

Creates an application on an existing, deployed cluster — defining its services, databases and environment variables — and then triggers a deployment.

  • Arguments: account_id, cluster_id, application_name (max 60 characters); optional environment (development, staging or production, default production), services, databases, environment_variables
  • Scope: application:post
  • Asynchronous. Returns an operation handle; poll it with get_operation.

Each entry in services[] needs a name and a source_type:

  • source_type: "image" — supply image.
  • source_type: "git" — supply git_repo, git_branch, build_root and dockerfile_path.

Services also accept command, desired_count, daemon_set, server_pool_ids, ports[] (internal_port, optional external_port, and protocol — http, https, tcp, tcp_over_tls or udp) and volume_mounts[] (mount_path, and either empty_dir or a storage_size_gb plus storage_class_name).

Each entry in databases[] either attaches an existing database by existing_database_id (from list_cluster_databases), or creates a new one with type, version, storage_size_gb, storage_class_name and optional server_pool_ids. Both forms accept migration_command and migration_service_name.

Operations

get_operation

Returns the status of an asynchronous operation. Use it to follow deployments, cluster builds, cloud registration and any other long-running work.

  • Arguments: account_id, operation_id
  • Scope: operation:get

Tools that return an operation handle tell the agent to poll this one, so in practice you'll see it used automatically whenever you ask for something that takes time.