shadeform
Find, launch and manage GPU instances, clusters, volumes, SSH keys and templates on Shadeform.
Setup guide and tool reference

Install

Select a method below to install the server.
Cursor
Claude Code
Claude Desktop
ChatGPT Desktop
VS Code
Antigravity CLI
Antigravity IDE
Codex CLI
opencode
Available Tools (24)
volumes_get Get details for the specified storage volume. ▾
Get details for the specified storage volume.
Read-only Idempotent Open-world
volumes_delete Delete a storage volume. This action is irreversible and destroys all data stored on the volume. Before calling this tool, restate the volume name and id to the user and wait for their explicit confirmation. Never delete a volume the user did not name. ▾
Delete a storage volume. This action is irreversible and destroys all data stored on the volume. Before calling this tool, restate the volume name and id to the user and wait for their explicit confirmation. Never delete a volume the user did not name.
Open-world
volumes_list_types Get list of supported storage volumes. ▾
Get list of supported storage volumes.
Read-only Idempotent Open-world
volumes_create Create a new storage volume ▾
Create a new storage volume
Open-world
volumes_list Get all storage volumes for the account. ▾
Get all storage volumes for the account.
Read-only Idempotent Open-world
templates_update Update an existing template ▾
Update an existing template
Idempotent Open-world
templates_get Get information about a specific template ▾
Get information about a specific template
Read-only Idempotent Open-world
templates_delete Delete a template. This action is irreversible. Before calling this tool, restate the template name and id to the user and wait for their explicit confirmation. ▾
Delete a template. This action is irreversible. Before calling this tool, restate the template name and id to the user and wait for their explicit confirmation.
Open-world
templates_save Create a new template ▾
Create a new template
Open-world
templates_list_featured List featured templates ▾
List featured templates
Read-only Idempotent Open-world
templates_list List all templates created by the user ▾
List all templates created by the user
Read-only Idempotent Open-world
ssh_keys_set_default Set the specified SSH Key as the default SSH Key used by Shadeform. ▾
Set the specified SSH Key as the default SSH Key used by Shadeform.
Open-world
ssh_keys_get Get details for the specified SSH Key in the url. ▾
Get details for the specified SSH Key in the url.
Read-only Idempotent Open-world
ssh_keys_delete Delete an SSH key. The Shadeform managed SSH key, the current default SSH key, and SSH keys in use by instances cannot be deleted. This action is irreversible. Before calling this tool, restate the key name and id to the user and wait for their explicit confirmation. ▾
Delete an SSH key. The Shadeform managed SSH key, the current default SSH key, and SSH keys in use by instances cannot be deleted. This action is irreversible. Before calling this tool, restate the key name and id to the user and wait for their explicit confirmation.
Open-world
ssh_keys_add Add a new SSH Key ▾
Add a new SSH Key
Open-world
ssh_keys_list Get all SSH Keys for the account. ▾
Get all SSH Keys for the account.
Read-only Idempotent Open-world
instances_update_model Queues a new model configuration for an instance launched to serve a model. The agent on the instance applies it, so a 200 means the change was accepted and returns the id of the queued entry. Only instances launched with a model configuration accept this. ▾
Queues a new model configuration for an instance launched to serve a model. The agent on the instance applies it, so a 200 means the change was accepted and returns the id of the queued entry. Only instances launched with a model configuration accept this.
Open-world
instances_update Update mutable details about the instance. Set a value to null to delete it. Omit a value or leave undefined to keep unchanged. ▾
Update mutable details about the instance. Set a value to null to delete it. Omit a value or leave undefined to keep unchanged.
Open-world
instances_restart Restart an instance. The status of the instance will stay as 'active' throughout, but you may have to wait a few minutes for the instance to be ready to use again. ▾
Restart an instance. The status of the instance will stay as 'active' throughout, but you may have to wait a few minutes for the instance to be ready to use again.
Open-world
instances_get Get details for the specified, non deleted, instance in the url. ▾
Get details for the specified, non deleted, instance in the url.
Read-only Idempotent Open-world
instances_delete Delete an instance. This will move the instance to the 'deleting' status while the instance is being deleted. Once the instance has entered the 'deleting' status, the account will no longer be billed for the instance. This action is irreversible and destroys any data on the instance's local disk. Before calling this tool, restate the instance name and id to the user and wait for their explicit confirmation. Never delete an instance the user did not name. ▾
Delete an instance. This will move the instance to the 'deleting' status while the instance is being deleted. Once the instance has entered the 'deleting' status, the account will no longer be billed for the instance. This action is irreversible and destroys any data on the instance's local disk. Before calling this tool, restate the instance name and id to the user and wait for their explicit confirmation. Never delete an instance the user did not name.
Open-world
instances_list_types Return all the GPU instance types with their corresponding availability and specs. Every `availability` entry carries a `rental_type` (`on_demand` or `spot`). Instance types that offer spot list each region twice, once per rental type, and the two entries can differ in `available`. If your integration keys on `region` or checks whether any region is available, filter to `rental_type: on_demand` first to keep the previous behavior. See the [Spot Instances guide](/guides/spotinstances) for details. ▾
Return all the GPU instance types with their corresponding availability and specs. Every `availability` entry carries a `rental_type` (`on_demand` or `spot`). Instance types that offer spot list each region twice, once per rental type, and the two entries can differ in `available`. If your integration keys on `region` or checks whether any region is available, filter to `rental_type: on_demand` first to keep the previous behavior. See the [Spot Instances guide](/guides/spotinstances) for details.
Read-only Idempotent Open-world
instances_create Create a new GPU instance. Our create API is designed to be asynchronous, so the response will be a CreateResponse object with a status of "creating". We then have a process that will pick it up and create it. You can poll the /instances/{id}/info endpoint to check the status of the instance. ▾
Create a new GPU instance. Our create API is designed to be asynchronous, so the response will be a CreateResponse object with a status of "creating". We then have a process that will pick it up and create it. You can poll the /instances/{id}/info endpoint to check the status of the instance.
Open-world
instances_list Get all non deleted instances. Note: instances in the "deleting" status will also show up here. ▾
Get all non deleted instances. Note: instances in the "deleting" status will also show up here.
Read-only Idempotent Open-world
No tools match your search.
Server Instructions
These instructions are provided to LLMs when they connect to help them understand how to use the server effectively.
Shadeform MCP server: find, launch and manage GPU instances, clusters, volumes, SSH keys and templates across Shadeform's cloud marketplace. Requests are authenticated with the user's Shadeform API key (header MCP-SHADEFORM-API-API-KEY-AUTH); all tools act on that account. Prices are USD per hour. Launching instances and clusters bills the account immediately; delete tools are irreversible, so always restate the resource name and id and wait for explicit confirmation before calling them. Instance creation is asynchronous: poll instances_get until status is active. Docs: https://docs.shadeform.ai/guides/mcp
Server URL
https://mcp.shadeform.ai
Raw Configuration
If you have a client that's not listed above, access the raw MCP configuration below. Check out the troubleshooting documentation for more help.
{ "command": "npx", "args": [ "mcp-remote@0.1.25", "https://mcp.shadeform.ai", "--header", "Mcp-Shadeform-Api-Api-Key-Auth:${MCP_SHADEFORM_API_API_KEY_AUTH}", "--header", "Mcp-Shadeform-Api-Oauth2-Access-Token:${MCP_SHADEFORM_API_OAUTH2_ACCESS_TOKEN}" ], "env": { "MCP_SHADEFORM_API_API_KEY_AUTH": "<your-value-here>", "MCP_SHADEFORM_API_OAUTH2_ACCESS_TOKEN": "<your-value-here>" } }