Partiri MCP server

MCP Server

Partiri Cloud provides a Model Context Protocol (MCP) server that lets AI coding assistants manage your services directly. You can create, deploy, monitor, and troubleshoot services without leaving your editor or terminal. The MCP server is hosted by Partiri — there is nothing to install or run yourself. You just configure your AI client to connect to it.

Setup

The quickest way to get started is with the Partiri CLI. Run partiri mcp install and select your AI client from the list. The CLI writes the correct configuration to your client's MCP config file automatically. You can also specify the client directly with partiri mcp install --client claude-code. To remove the configuration later, run partiri mcp uninstall. You need a Partiri API key to authenticate — if you haven't set one up yet, run partiri auth login first.

HTTP (remote)

Configure your MCP client to connect to the hosted endpoint:

{
  "partiri-cloud": {
    "type": "http",
    "url": "https://mcp.partiri.cloud/mcp"
  }
}

No API key is needed in the config — the OAuth 2.1 browser flow handles authentication. When you first connect, your AI client will open a browser window where you enter your Partiri API key.

stdio (local)

For local use, set your API key as an environment variable and point the client at the binary:

{
  "partiri-cloud": {
    "type": "stdio",
    "command": "node",
    "args": ["dist/index.mjs"],
    "env": { "PARTIRI_API_KEY": "your-key" }
  }
}

The API key is resolved from the PARTIRI_API_KEY environment variable, or from the file ~/.config/partiri/key written by partiri auth login.

Authentication

The HTTP transport implements OAuth 2.1 with authorization code + PKCE. When a client connects for the first time, it performs the following flow:

Client discovers endpoints via /.well-known/oauth-protected-resource/mcp
Client dynamically registers via POST /register
A browser window opens the /authorize page where you enter your Partiri API key
The key is validated and an encrypted authorization code is issued
The code is exchanged for access and refresh tokens
All subsequent requests use Authorization: Bearer <token>

Tokens are stateless — your API key is AES-256-GCM encrypted inside the token itself. No token database is required. Legacy x-api-key header authentication is also supported for backward compatibility.

Available tools

The MCP server exposes 23 tools organized by domain. Workspaces and user: list_workspaces, get_current_user. Projects and resources: list_projects, create_project, list_pods, list_regions. Pricing and billing: get_pricing, get_balance. Services: list_services, get_service, create_service, update_service, validate_service. Deployments and control: deploy_service, pause_service, unpause_service, list_jobs. Storage: list_volumes, get_volume. Monitoring: get_cpu_metrics, get_memory_metrics, get_network_metrics. The use_partiri_cli tool returns guidance for running the partiri CLI yourself for sensitive operations — secrets, volume changes, environment variables, and service deletion — that are intentionally not exposed as direct MCP tools.

Tool	Description
`list_workspaces`	List all workspaces the user has access to
`get_current_user`	Get the authenticated user's profile
`list_projects`	List projects in a workspace
`create_project`	Create a new project
`list_pods`	List available compute pods
`list_regions`	List available deployment regions
`get_pricing`	Get pod and volume pricing for a region
`get_balance`	Get the current workspace balance
`list_services`	List services in a project
`get_service`	Get full service configuration (environment variables omitted)
`create_service`	Create a new service
`update_service`	Update service configuration
`validate_service`	Preflight-validate a service configuration before create or update
`deploy_service`	Trigger a new deployment
`pause_service`	Pause a running service
`unpause_service`	Resume a paused service
`list_jobs`	List deployment jobs for a service
`list_volumes`	List persistent volumes in a project
`get_volume`	Get details of a single persistent volume
`get_cpu_metrics`	Get CPU usage metrics
`get_memory_metrics`	Get memory usage metrics
`get_network_metrics`	Get network metrics
`use_partiri_cli`	Guidance for running the partiri CLI for CLI-only operations (secrets, volume changes, env vars, deletion)

Supported clients

The CLI can configure the following AI clients automatically: Claude Desktop, Claude Code, Cursor, VS Code (GitHub Copilot), Copilot CLI, and Windsurf. The install command detects the correct config file path for your operating system (Linux, macOS, and Windows) and writes the Partiri MCP server entry. If your client is not listed, you can add the configuration manually by adding a partiri-cloud entry pointing to https://mcp.partiri.cloud/mcp in your client's MCP settings file.

Usage examples

Inspect a service and check resource usage

Prompt your AI agent:

Show me CPU and memory usage for my api service over the last hour.

The agent resolves the workspace and project, looks up the service, then fetches CPU and memory metrics.

Deploy a service and monitor progress

Prompt your AI agent:

Deploy the latest commit of the frontend service in my production project, then tell me when it succeeds.

The agent triggers a deployment and polls the job list until it completes.

Create a service from a Git repository

Prompt your AI agent:

List all services in my staging project and create a new Node.js web service from github.com/owner/repo on the main branch.

The agent lists existing services, selects an appropriate pod and region, then creates the new service linked to the repository.

Support

For questions, bug reports, or integration help: [email protected]