CP MCP Server

URL: https://mcp.computeportal.io/mcp

The ComputePortal MCP server exposes all CP platform capabilities as tools that AI assistants (Claude, Cursor, Windsurf, and any MCP-compatible client) can call directly. This lets you manage GPU workloads, deploy services, attach domains, and inspect logs through natural language — without leaving your AI assistant.

Protocol

The CP MCP server implements the Model Context Protocol (MCP) using the Streamable HTTP transport. It is built on the @modelcontextprotocol/sdk at version 1.29.0.

Property	Value
Protocol	MCP (Model Context Protocol)
Transport	Streamable HTTP
Endpoint	`https://mcp.computeportal.io/mcp`
SDK	`@modelcontextprotocol/sdk` 1.29.0
Auth	`CP_API_TOKEN` environment variable

Authentication

The MCP server authenticates using your CP API token, set as an environment variable in your client configuration:

CP_API_TOKEN=cp_live_<hex-string>

The token format is cp_live_ followed by a lowercase hex string. This is the same token used for direct API calls. Find your token at https://dashboard.computeportal.io under Settings → API Tokens.

Integrations

Claude Desktop

Add the following to your Claude Desktop configuration file.

macOS: ~/Library/Application Support/Claude/claude_desktop_config.jsonWindows: %APPDATA%\Claude\claude_desktop_config.json

json

{
  "mcpServers": {
    "computeportal": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.computeportal.io/mcp"],
      "env": {
        "CP_API_TOKEN": "cp_live_4a7f3b9c2e1d8a06"
      }
    }
  }
}

After saving the file, restart Claude Desktop. You will see a hammer icon in the chat input area indicating MCP tools are available.

Verify the connection: Ask Claude "List my CP services" — it should call cp_service_list and return your running services.

Cursor

Open Cursor Settings → MCP and add a new server entry:

json

{
  "computeportal": {
    "command": "npx",
    "args": ["-y", "mcp-remote", "https://mcp.computeportal.io/mcp"],
    "env": {
      "CP_API_TOKEN": "cp_live_4a7f3b9c2e1d8a06"
    }
  }
}

Or, if your Cursor version supports HTTP MCP servers natively, configure it directly:

json

{
  "computeportal": {
    "url": "https://mcp.computeportal.io/mcp",
    "headers": {
      "Authorization": "Bearer cp_live_4a7f3b9c2e1d8a06"
    }
  }
}

Windsurf

In your Windsurf MCP settings (usually ~/.windsurf/mcp.json):

json

{
  "servers": {
    "computeportal": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.computeportal.io/mcp"],
      "env": {
        "CP_API_TOKEN": "cp_live_4a7f3b9c2e1d8a06"
      }
    }
  }
}

Direct HTTP (Advanced)

If your client supports Streamable HTTP transport natively, you can connect directly without mcp-remote:

URL:     https://mcp.computeportal.io/mcp
Headers: Authorization: Bearer cp_live_4a7f3b9c2e1d8a06
         Content-Type: application/json

For clients that use the MCP initialize handshake, send:

bash

curl -X POST https://mcp.computeportal.io/mcp \
  -H "Authorization: Bearer cp_live_4a7f3b9c2e1d8a06" \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "initialize",
    "params": {
      "protocolVersion": "2024-11-05",
      "capabilities": {},
      "clientInfo": { "name": "my-client", "version": "1.0.0" }
    }
  }'

Tools Reference

The CP MCP server exposes 21 tools across domains, services, environment, logs, and scaling.

Domain Tools

`cp_domain_add`

Attach a custom hostname to a CP service. Validates the DNS CNAME record and provisions nginx + Let's Encrypt certificate automatically.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name to attach the domain to (e.g. `cp-dashboard`)
`hostname`	string	Yes	The fully-qualified hostname (e.g. `dashboard.example.com`)
`skip_dns`	boolean	No	Skip DNS CNAME validation. Default: `false`

Example prompt

"Attach dashboard.example.com to my cp-dashboard service"

Claude will call:

json

{
  "tool": "cp_domain_add",
  "arguments": {
    "service": "cp-dashboard",
    "hostname": "dashboard.example.com"
  }
}

Returns — the new domain record:

json

{
  "hostname": "dashboard.example.com",
  "status": "active",
  "service": "cp-dashboard",
  "created_at": "2026-07-04T10:52:36.153Z"
}

`cp_domain_list`

List all hostnames attached to a CP service, including the default <service>.cpctl.app domain.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name (e.g. `cp-dashboard`)

Example prompt

"What domains are attached to cp-dashboard?"

Returns

json

{
  "domains": [
    { "hostname": "cp-dashboard.computeportal.io", "status": "active", "created_at": null },
    { "hostname": "dashboard.example.com", "status": "active", "created_at": "2026-07-04T10:52:36.153Z" }
  ]
}

`cp_domain_remove`

Remove a custom hostname from a CP service. Deletes the nginx virtual host and TLS certificate. The default <service>.cpctl.app domain cannot be removed.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name (e.g. `cp-dashboard`)
`hostname`	string	Yes	The hostname to remove (e.g. `dashboard.example.com`)

Example prompt

"Remove dashboard.example.com from cp-dashboard"

Returns — confirmation that the domain was removed.

Service Tools

`cp_service_list`

List all services deployed in your CP environment, including their status, image, and replica count.

Inputs

Parameter	Type	Required	Description
`environment`	string	No	Filter by environment name (e.g. `production`)

Example prompt

"Show me all my running services"

Claude will call:

json

{
  "tool": "cp_service_list",
  "arguments": {}
}

`cp_service_get`

Get details for a specific service — image, replicas, ports, resource limits, and current status.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name (e.g. `ml-inference`)

Example prompt

"Show me the details for ml-inference"

`cp_deploy`

Deploy a new service from a Docker image or a public GitHub repo. When a repo URL is provided, CP clones it and builds the Docker image internally from the repo's Dockerfile — no registry setup required.

Inputs

Parameter	Type	Required	Description
`image`	string	No*	Docker image to deploy (e.g. `myorg/agent:latest`). Provide either `image` or `repo`.
`repo`	string	No*	Public GitHub repo URL (e.g. `https://github.com/org/repo`). CP builds the image from the `Dockerfile` internally. Provide either `repo` or `image`.
`name`	string	No	Service name (default: derived from image or repo name)
`region`	string	No	Region ID (e.g. `eu-west`)
`gpu`	string	No	GPU type: `rtx4090` \| `rtx5090`
`gpu_count`	integer	No	Number of GPUs
`env`	object	No	Environment variables as key-value pairs
`domains`	array	No	Custom hostnames to attach after deploy

* One of image or repo is required.

Example — deploy from a Docker image

"Deploy a new service called ml-inference using the image registry.example.com/model:v2"

Claude will call:

json

{
  "tool": "cp_deploy",
  "arguments": {
    "name": "ml-inference",
    "image": "registry.example.com/model:v2"
  }
}

Example — deploy from a GitHub repo

"Deploy https://github.com/my-org/my-app as a new service called my-app"

Claude will call:

json

{
  "tool": "cp_deploy",
  "arguments": {
    "name": "my-app",
    "repo": "https://github.com/my-org/my-app"
  }
}

CP clones the repo, builds the image from the Dockerfile, pushes it to an internal registry, and starts the service. The build typically takes 1–3 minutes. Only public repositories are supported — for private repo access, contact Compute Portal.

Returns

json

{
  "service_id": "my-app",
  "name": "my-app",
  "url": "https://my-app.cpctl.app",
  "status": "running"
}

`cp_service_update`

Update a service's image, resource limits, or configuration. Triggers a rolling update.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name to update
`image`	string	No	New container image reference
`replicas`	integer	No	New replica count

Example prompt

"Update ml-inference to use image registry.example.com/model:v3"

`cp_service_delete`

Remove a service and all associated resources from the CP platform.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name to delete

Example prompt

"Delete the ml-inference service"

`cp_service_scale`

Scale a service to a target number of replicas.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name to scale
`replicas`	integer	Yes	Target replica count

Example prompt

"Scale ml-inference to 3 replicas"

Claude will call:

json

{
  "tool": "cp_service_scale",
  "arguments": {
    "service": "ml-inference",
    "replicas": 3
  }
}

`cp_service_restart`

Force a rolling restart of all service replicas without changing the image or configuration.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name to restart

Example prompt

"Restart ml-inference"

`cp_service_logs`

Stream or fetch recent log output from a service.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name
`tail`	integer	No	Number of most recent log lines to return. Default: `100`

Example prompt

"Show me the last 100 lines of logs from ml-inference"

Claude will call:

json

{
  "tool": "cp_service_logs",
  "arguments": {
    "service": "ml-inference",
    "tail": 100
  }
}

`cp_service_inspect`

Return the raw Docker Swarm service spec for a service — useful for debugging configuration.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name to inspect

Example prompt

"Show me the raw spec for ml-inference"

`cp_env_list`

List all environment variables set on a service.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name

Example prompt

"What environment variables are set on cp-dashboard?"

`cp_env_set`

Set one or more environment variables on a service. Triggers a rolling restart to apply changes.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name
`vars`	object	Yes	Key-value map of variables to set (e.g. `{"MODEL_PATH": "/models/v2"}`)

Example prompt

"Set MODEL_PATH=/models/v2 and BATCH_SIZE=32 on ml-inference"

Claude will call:

json

{
  "tool": "cp_env_set",
  "arguments": {
    "service": "ml-inference",
    "vars": {
      "MODEL_PATH": "/models/v2",
      "BATCH_SIZE": "32"
    }
  }
}

`cp_env_unset`

Remove one or more environment variables from a service. Triggers a rolling restart.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name
`keys`	array	Yes	List of variable names to remove (e.g. `["DEBUG", "OLD_KEY"]`)

Example prompt

"Remove the DEBUG variable from ml-inference"

`cp_gpu_list`

List available GPU resources on the CP DGX Spark platform, including availability and specs.

Inputs

Parameter	Type	Required	Description
`region`	string	No	Filter by region (e.g. `us-east-1`)

Example prompt

"What GPU resources are available?"

`cp_gpu_status`

Show current GPU utilization and allocation per service.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name to check GPU status for

Example prompt

"Show me GPU usage for ml-inference"

`cp_volume_list`

List persistent volumes attached to a service.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name

Example prompt

"What volumes are attached to ml-inference?"

`cp_volume_create`

Create a new persistent volume and attach it to a service.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name to attach the volume to
`path`	string	Yes	Mount path inside the container (e.g. `/data`)
`size`	string	No	Volume size (e.g. `10Gi`). Default: `5Gi`

Example prompt

"Create a 10Gi volume at /data for ml-inference"

Claude will call:

json

{
  "tool": "cp_volume_create",
  "arguments": {
    "service": "ml-inference",
    "path": "/data",
    "size": "10Gi"
  }
}

`cp_volume_delete`

Remove a persistent volume from a service. This action is irreversible and deletes all data in the volume.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name the volume is attached to
`name`	string	Yes	The volume name to delete

Example prompt

"Delete the ml-data volume from ml-inference"

`cp_health_check`

Check the health status of a service and its replicas — useful for diagnosing crash loops or degraded deployments.

Inputs

Parameter	Type	Required	Description
`service`	string	Yes	The service name to check

Example prompt

"Is ml-inference healthy? How many replicas are running?"

Claude will call:

json

{
  "tool": "cp_health_check",
  "arguments": {
    "service": "ml-inference"
  }
}

Returns

json

{
  "service": "ml-inference",
  "status": "healthy",
  "replicas": { "running": 3, "desired": 3 },
  "last_updated": "2026-07-08T09:12:00.000Z"
}

Usage Patterns

Natural Language Examples

The AI assistant interprets natural language and maps it to the appropriate tool calls. You do not need to know tool names — just describe what you want.

Service management

"Show me all my running services and their GPU usage" → calls cp_service_list then cp_gpu_list

"Deploy a new service called ml-inference using the image registry.example.com/model:v2" → calls cp_deploy with image

"Deploy https://github.com/my-org/my-app as my-app" → calls cp_deploy with repo — CP builds the image from the Dockerfile automatically

"Scale ml-inference to 3 replicas" → calls cp_scale

Domains

"Attach api.myapp.com to my ml-inference service — I've already set up the CNAME" → calls cp_domain_add

"Show me all the domains on cp-dashboard" → calls cp_domain_list

Debugging

"Show me the last 100 lines of logs from ml-inference" → calls cp_service_logs

"My service is crashing — check its health and show me recent errors" → calls cp_health_check then cp_service_logs

Environment variables

"Set MODEL_PATH=/models/v2 and BATCH_SIZE=32 on ml-inference" → calls cp_env_set

"What environment variables are set on cp-dashboard?" → calls cp_env_list

Troubleshooting

"No tools available" in Claude Desktop

Verify your claude_desktop_config.json is valid JSON (no trailing commas, matching brackets)
Confirm the CP_API_TOKEN value starts with cp_live_
Fully quit and relaunch Claude Desktop (Cmd+Q on macOS, not just close the window)
Check that npx is in your PATH: run which npx in a terminal

"Unauthorized" errors when calling tools

Your token may be invalid or expired. Generate a new token at https://dashboard.computeportal.io under Settings → API Tokens, then update your MCP configuration.

DNS validation failure when adding a domain

The cp_domain_add tool validates that your hostname's CNAME record resolves to cpctl.app. If the check fails:

Confirm your DNS record is saved and set to DNS only (not proxied)
Wait for DNS propagation — TTL-dependent, typically 1–5 minutes for Cloudflare, up to 48 hours for some registrars
Verify with: dig CNAME dashboard.example.com — the answer should show cpctl.app
As a last resort, pass skip_dns: true to the tool and attach the domain immediately

Tool calls time out

The MCP server performs real infrastructure operations (certbot, nginx reload, Docker Swarm updates). Some operations take 10–30 seconds. This is normal. If a call times out in your client, check the CP Dashboard to see if the operation completed.

Security

Tokens are scoped to your CP account. Never share your cp_live_xxx token.
The MCP server does not store tokens — it passes them directly to the CP API on each tool call.
All traffic is over HTTPS. The MCP endpoint does not support plain HTTP.
Revoke tokens at any time from the CP Dashboard.

Reference

Item	Value
MCP server URL	`https://mcp.computeportal.io/mcp`
Protocol	MCP over Streamable HTTP
SDK	`@modelcontextprotocol/sdk` 1.29.0
Auth env var	`CP_API_TOKEN`
Token format	`cp_live_<hex>`
Total tools	21
API base URL	`https://api.computeportal.io`

Quickstart Guide — get a token, make your first API call, attach a custom domain
CP API Reference — full HTTP API documentation
CP Dashboard — manage everything visually
MCP Protocol Specification — official MCP documentation

CP MCP Server ​

Protocol ​

Authentication ​

Integrations ​

Claude Desktop ​

Cursor ​

Windsurf ​

Direct HTTP (Advanced) ​

Tools Reference ​

Domain Tools ​

cp_domain_add ​

cp_domain_list ​

cp_domain_remove ​

Service Tools ​

cp_service_list ​

cp_service_get ​

cp_deploy ​

cp_service_update ​

cp_service_delete ​

cp_service_scale ​

cp_service_restart ​

cp_service_logs ​

cp_service_inspect ​

cp_env_list ​

cp_env_set ​

cp_env_unset ​

cp_gpu_list ​

cp_gpu_status ​

cp_volume_list ​

cp_volume_create ​

cp_volume_delete ​

cp_health_check ​

Usage Patterns ​

Natural Language Examples ​

Troubleshooting ​

"No tools available" in Claude Desktop ​

"Unauthorized" errors when calling tools ​

DNS validation failure when adding a domain ​

Tool calls time out ​

Security ​

Reference ​

Related ​

CP MCP Server

Protocol

Authentication

Integrations

Claude Desktop

Cursor

Windsurf

Direct HTTP (Advanced)

Tools Reference

Domain Tools

`cp_domain_add`

`cp_domain_list`

`cp_domain_remove`

Service Tools

`cp_service_list`

`cp_service_get`

`cp_deploy`

`cp_service_update`

`cp_service_delete`

`cp_service_scale`

`cp_service_restart`

`cp_service_logs`

`cp_service_inspect`

`cp_env_list`

`cp_env_set`

`cp_env_unset`

`cp_gpu_list`

`cp_gpu_status`

`cp_volume_list`

`cp_volume_create`

`cp_volume_delete`

`cp_health_check`

Usage Patterns

Natural Language Examples

Troubleshooting

"No tools available" in Claude Desktop

"Unauthorized" errors when calling tools

DNS validation failure when adding a domain

Tool calls time out

Security

Reference

Related