loading…
Search for a command to run...
loading…
An MCP server for the Spot.io API that enables management of AWS and Azure Ocean clusters across multiple accounts. It provides tools for cluster inventory, nod
An MCP server for the Spot.io API that enables management of AWS and Azure Ocean clusters across multiple accounts. It provides tools for cluster inventory, node management, cost analysis, and scaling operations through natural language.
PyPI version CI License: MIT Python 3.10+
MCP server for the Spot.io (Spotinst) API. Supports both AWS and Azure Ocean clusters with multi-account access.
| Tool | Description |
|---|---|
list_accounts |
List all Spotinst accounts accessible with the current token |
list_all_clusters |
List ALL clusters across ALL accounts and clouds (AWS + Azure) |
| Tool | Description |
|---|---|
list_clusters |
List AWS Ocean K8s clusters |
get_cluster |
Get AWS Ocean cluster details |
list_vngs |
List AWS Virtual Node Groups (launch specs) |
get_vng |
Get AWS VNG details |
list_elastigroups |
List all Elastigroups |
get_elastigroup |
Get Elastigroup details |
get_allowed_instance_types |
Get allowed EC2 instance types |
get_right_sizing |
Get right-sizing resource suggestions (AWS only) |
| Tool | Description |
|---|---|
list_clusters_azure |
List Azure Ocean clusters |
get_cluster_azure |
Get Azure Ocean cluster details |
list_vngs_azure |
List Azure Virtual Node Groups |
get_vng_azure |
Get Azure VNG details |
cloud="azure" for Azure)| Tool | Description |
|---|---|
get_cluster_nodes |
List nodes in an Ocean cluster |
get_cluster_costs |
Get aggregated cost breakdown by namespace or resource |
list_rolls |
List deployment rolls |
get_roll |
Get roll details |
get_cluster_log |
Get scaling and activity log events |
| Tool | Description |
|---|---|
list_stateful_nodes |
List all Stateful Nodes (Managed Instances) |
get_stateful_node |
Get Stateful Node details |
| Tool | Description |
|---|---|
get_cluster_scheduling |
Get scheduling and auto-scaler configuration |
get_cluster_health |
Composite health check: nodes, recent errors, active rolls |
| Tool | Description |
|---|---|
get_cost_trending |
Week-over-week (or custom period) cost comparison |
get_savings_summary |
30-day cost and savings summary |
| Tool | Description |
|---|---|
filter_clusters_by_tag |
Filter clusters by tag key/value |
filter_vngs_by_tag |
Filter VNGs by tag key/value |
| Tool | Description |
|---|---|
export_cluster_yaml |
Export cluster config as YAML for GitOps/backup |
export_vng_yaml |
Export VNG config as YAML for GitOps/backup |
confirm=true)| Tool | Description |
|---|---|
remove_instances |
Recommended — Remove instances using a named strategy (see below) |
initiate_roll |
Rolling restart of nodes in an Ocean cluster |
detach_instances |
Detach and optionally terminate instances from an AWS Ocean cluster |
update_vng |
Update an AWS VNG configuration |
update_vng_azure |
Update an Azure VNG configuration |
All tools accept an optional account_id parameter to query any account.
The remove_instances tool provides a safe, intent-based interface for removing instances. Instead of remembering which API flags to set, you pick a strategy:
| Strategy | What happens | Use case |
|---|---|---|
drain_and_replace |
Gracefully drain pods (respects PDBs), terminate, Ocean replaces. Default and safest. | Replacing a problematic node in production |
replace |
Immediately terminate, Ocean auto-replaces. No graceful drain. AWS only. | Fast replacement when drain isn't needed |
remove_permanently |
Terminate + reduce cluster capacity. No replacement. AWS only. | Downsizing the cluster |
When confirm=false (default), the tool shows a detailed execution plan so you can review before proceeding:
remove_instances("o-abc123", "i-abc123", strategy="drain_and_replace")
→ SAFETY: Action NOT executed. Set confirm=true to proceed.
DRAIN AND REPLACE 1 instance(s) in cluster o-abc123:
Instances: ['i-abc123']
Method: Rolling restart (20% per batch)
- Pods will be gracefully drained (PDBs respected)
- Instances will be terminated after drain
- Ocean will automatically launch replacements
export SPOTINST_TOKEN="your-spotinst-api-token"
export SPOTINST_ACCOUNT_ID="act-xxxxxxxx"
pip install mcp-server-spotinst
uvx mcp-server-spotinst
Add to ~/.mcp.json:
{
"mcpServers": {
"spotinst": {
"command": "uvx",
"args": ["mcp-server-spotinst"],
"env": {
"SPOTINST_TOKEN": "your-token",
"SPOTINST_ACCOUNT_ID": "act-xxxxxxxx"
}
}
}
}
docker build -t mcp-server-spotinst .
docker run -e SPOTINST_TOKEN=your-token -e SPOTINST_ACCOUNT_ID=act-xxxxxxxx mcp-server-spotinst
mcp-server-spotinst
Your personal API token can access multiple Spotinst accounts. Use list_accounts to see all available accounts, then pass account_id to any tool:
list_clusters(account_id="act-be5e7ffe")
get_cluster_nodes("o-390ef886", account_id="act-9785011e", cloud="azure")
Or use list_all_clusters for a single-call inventory across everything.
Uses the Spot.io REST API at https://api.spotinst.io.
Добавь это в claude_desktop_config.json и перезапусти Claude Desktop.
{
"mcpServers": {
"spotinst-mcp-server": {
"command": "npx",
"args": []
}
}
}