cloud-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CLOUD_TIMEOUT | No | HTTP timeout in seconds. | 60 |
| TIMEWEB_TOKEN | No | Legacy Timeweb token for compatibility with the old timeweb-mcp; used only if the cloud accounts file does not exist. | |
| CLOUD_READONLY | No | Set to 1 to prohibit all operations except read operations. | 0 |
| CLOUD_CACHE_DIR | No | Directory where cloud_catalog writes its files. | ~/.cache/cloud-mcp |
| CLOUD_PROVIDERS | No | Comma-separated subset of providers to enable, e.g. timeweb,selectel. | all |
| CLOUD_SPECS_DIR | No | Custom directory for API snapshot specifications. | |
| SELECTEL_REGION | No | Region used for regional Selectel products. | ru-2 |
| CLOUD_ACCOUNTS_FILE | No | Path to the accounts file containing credentials for all providers. | ~/.config/cloud-mcp/accounts.json |
| SELECTEL_BASE_<SLUG> | No | Point override for a specific Selectel product base URL, e.g. SELECTEL_BASE_MANAGED_DATABASES. | |
| TIMEWEB_ACCOUNTS_FILE | No | Legacy Timeweb accounts file for compatibility with the old timeweb-mcp; used only if the cloud accounts file does not exist. | |
| CLOUD_ALLOW_DESTRUCTIVE | No | Set to 1 to allow destructive/paid operations when CLOUD_READONLY=0. | 0 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
Multi-provider cloud MCP (timeweb, selectel, tilda). Discovery-first:
cloud_search(query, provider=...) -> ranked operations;
cloud_describe(provider, tool) -> full parameter/body schema;
cloud_call(provider, tool, args) -> execute (read runs freely; write needs confirm=true; destructive/paid also need CLOUD_READONLY=0 + CLOUD_ALLOW_DESTRUCTIVE=1). Wide overview: cloud_catalog(provider) writes a greppable file with every operation. Errors: '[code] message'.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| cloud_searchA | Find API operations across providers by a free-form query (RU or EN); returns the best matches with their tool names, methods and parameters |
| cloud_describeC | Full parameter and body schema of one operation |
| cloud_callC | Execute an operation. Flat args are routed to path/query/header/body by the operation schema; 'confirm': true is required for write/destructive/paid |
| cloud_providersB | Providers, their products, operation counts and credential state |
| cloud_catalogC | Write a greppable catalog file (rules + one line per operation) for a provider |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 5 tools
Each tool targets a distinct step: search discovers operations, describe returns one operation's schema, call executes, providers lists accounts/credentials, catalog writes a file. The only mild overlap is between cloud_search (find operations) and cloud_catalog (enumerate operations into a file), but the output-medium distinction is stated clearly enough.
All names share a consistent cloud_ prefix and snake_case style. However the suffixes mix verbs (search, describe, call) with nouns (providers, catalog), so the verb_noun pattern isn't strictly uniform, though it remains predictable and readable.
Five tools is a tight, well-scoped set for a generic multi-provider API gateway: discover, inspect, invoke, list context, and export. No tool feels redundant and there is no obvious missing piece at this layer.
The search→describe→call pipeline covers the full discovery-to-execution lifecycle, and providers/catalog fill in context and offline grep workflows. Minor gaps remain: no way to set/refresh credentials (only view state) and no built-in pagination or batch handling, though agents can work around these.