cloudhpc-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CLOUDHPC_APIKEY | No | Your cloudHPC API key. If not set, the server will look for a key in ~/.cfscloudhpc/apikey. |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| list_solversA | List the solvers/scripts available on cloudHPC, grouped by family. search: optional case-insensitive substring (e.g. "openfoam", "fds6.9"). |
| list_machine_optionsA | List the vCPU counts and RAM/instance types that can be requested. |
| suggest_resourcesA | Suggest vCPU and RAM type for a run, following cloudHPC scalability rules. solver: script name (e.g. "fds6.9.1", "openFoam-v2406") or family ("fds", "openfoam", "calculix", "code_aster", "openradioss", "su2"). cells: CFD cells (OpenFOAM/SU2). nodes: FEA nodes (CalculiX/code_aster). elements: OpenRadioss elements. fds_*: FDS mesh info if the .fds file cannot be read directly. case_folder: (local mode) folder to inspect automatically instead. prefer_speed: favour the faster hypercore/hypercpu instances over cost. |
| list_storageA | List files and folders in the user's cloudHPC storage. folder: storage folder path ("" = root, e.g. "myCase" or "myCase/sub"). |
| list_resultsA | List the result archives of a case folder (FDS.tar.gz, OPENFOAM-*.tar.gz, ...). |
| get_download_linkA | Get a temporary download link for a storage file (e.g. "myCase/FDS.tar.gz"). |
| get_upload_linkA | Get a signed URL to upload one file into a storage folder (remote clients). For a case folder: compress the CONTENT of the folder (files at the archive root, no wrapping folder) into e.g. upload.tar.gz or case.zip, then upload it into a storage folder with the case name. Accepted: zip, tar.gz, 7z, rar, xz. |
| delete_storageA | Delete a file or folder from storage. Irreversible. Call first with confirm=false to get the summary, show it to the user, and call again with confirm=true only after the user explicitly agrees. |
| launch_simulationA | Launch a simulation on a case folder already in storage. Costs money. solver: exact script name from list_solvers (e.g. "fds6.9.1"). cpu / ram: from suggest_resources or list_machine_options. folder: storage folder with the case. mesh_folder: optional storage folder with a mesh to reuse (OpenFOAM). regular_instance: non-preemptible machine (more expensive, no interruptions). Call first with confirm=false, show the returned summary to the user, and call again with confirm=true only after explicit approval. |
| list_simulationsB | List the user's simulations, most recent first. status: "active" (pending/running/stopping), "all", "completed", "error", "stopped". |
| get_simulationA | Get status and details of a simulation. include_log: add the last output/log lines and a diagnosis of known cloudHPC errors and warnings with their fix. Use it whenever a run ends, since a run can be COMPLETED even if the solver failed. |
| wait_for_simulationA | Wait until a simulation is no longer pending/running, or until max_minutes. Polls every poll_seconds (min 60, to respect API rate limits). If it returns still running, call it again later rather than looping quickly. |
| sync_simulationA | Ask a running simulation to upload partial results to storage now. |
| stop_simulationA | Stop a running simulation. soft: the solver stops cleanly and results are saved (recommended). hard: immediate termination. Call first with confirm=false and show the summary; confirm=true only after the user agrees. |
| open_remote_desktopA | Get the browser remote-desktop (VNC) link of a running simulation. The link is short-lived and personal: give it only to the user. |
| api_usageA | Show the API rate limits and how many calls have been used. |
| inspect_caseA | Inspect a local case folder: detect the solver and the model size. Reads .fds meshes (cells, MPI_PROCESS groups), OpenFOAM polyMesh cells, CalculiX nodes, code_aster mpi_nbcpu, and runs pre-flight checks for the most common cloudHPC errors (folder name, MPI_PROCESS order, decomposeParDict, missing files). Makes no API calls. Fix 'error' items before uploading. |
| upload_folderA | Compress a local case folder and upload it to cloudHPC storage. The CONTENT of the folder goes into upload.tar.gz, uploaded into the storage folder storage_folder (default: the local folder name). Hidden files are skipped. Returns the storage folder to use in launch_simulation. |
| download_resultsA | Download result archives of a storage folder to a local directory. folder: storage folder of the case. local_dir: where to save (created if missing). extract: unpack .tar.gz archives and delete them after. files_to_get: optional list of file names; default = all result archives. |
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 19 tools
Each tool targets a distinct action or resource: listing, launching, monitoring, stopping, syncing, uploading, downloading, storage management, and resource suggestion. Related storage and simulation tools are clearly separated by their descriptions.
The vast majority of tools follow a clear verb_noun snake_case pattern such as list_solvers, launch_simulation, and delete_storage. The only minor deviation is api_usage, which is a noun phrase rather than an imperative verb.
At 19 tools, the server is slightly above the typical well-scoped range, but every tool maps to a necessary part of the cloudHPC workflow: local inspection, upload, launch, monitoring, storage, and result retrieval. The count feels justified rather than bloated.
The tool set covers the full simulation lifecycle: inspect locally, upload, suggest resources, launch, monitor, wait, sync partial results, stop, access remote desktop, download results, and manage storage. There are no obvious dead ends or missing critical operations.