Model Context Protocol (MCP)

Build and Manage Game Server Infrastructure With AI.

PingCore has an MCP server. Connect it to Cursor, ChatGPT, Claude, or any compatible client. Generate integration code, follow workflow guides, deploy servers, and manage your fleet.

Connect Now API Documentation
MCP Endpointhttps://pingcore.io/mcp

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants connect to external tools and data sources. No more copy-pasting API docs into chat. The AI reads them on its own and can execute real actions too.

PingCore is one of the first game server platforms with native MCP support. One connection gives the AI everything it needs to build integrations using real API documentation and manage your infrastructure through natural language.

Cursor
Claude Desktop
ChatGPT
Windsurf
VS Code + Copilot
Any MCP Client
Your AI Assistant
Cursor, ChatGPT, Claude
↓ MCP
PingCore MCP Server
https://pingcore.io/mcp
↓ API
Build
Docs, code gen, guides
Manage
Deploy, scale, monitor

Connect in Seconds

One URL. Any MCP-compatible client. No SDK, no library, no setup script.

Cursor / VS Code

IDE with native MCP support

Add PingCore as an MCP server in your project or global config. Create a .cursor/mcp.json file in your project root:

{
"mcpServers": {
"pingcore": {
"url": "https://pingcore.io/mcp"
}
}
}

Restart Cursor. The PingCore tools will appear in agent mode automatically.

ChatGPT

Add as a connected app

Link PingCore inside ChatGPT as a connected app. After that you can build integrations or manage servers from any conversation.

  1. 1Open ChatGPT and go to Settings → Connected Apps
  2. 2Search for Pingcore or add by URL
  3. 3Authorize with your PingCore account
  4. 4Ask ChatGPT to manage your game servers
MCP Server URL:
https://pingcore.io/mcp

Claude Desktop

Anthropic's AI assistant

Add PingCore to your Claude Desktop configuration file.

// claude_desktop_config.json
{
"mcpServers": {
"pingcore": {
"url": "https://pingcore.io/mcp"
}
}
}

Any MCP Client

Open protocol, open access

MCP is an open standard. Windsurf, VS Code with Copilot agent mode, custom scripts, your own AI tools. If it speaks MCP, it connects to PingCore.

// Point any MCP client at:
https://pingcore.io/mcp
// Transport: Streamable HTTP (POST)
// Auth: OAuth 2.1 (auto-negotiated)
Build

Build Integrations Faster

The AI has access to PingCore's full API surface: documentation, schemas, and workflow guides. Generate working code without leaving your editor.

Search & Browse API Docs

The AI sees a typed schema for every tool your account can use, with its parameters. Ask "how do I deploy servers?" and it finds the right tools and the workflow guide for the task.

Generate Integration Code

Describe what you need. "A script that deploys Minecraft servers across EU." The AI writes working code in your language with the correct endpoints and auth patterns.

Follow Workflow Playbooks

Built-in guides cover deploying servers, provisioning users, managing resources, and accessing container APIs. The AI follows these playbooks and makes the right API calls in the right order.

Manage

Manage Infrastructure Conversationally

Beyond documentation. Deploy servers, monitor your fleet, and resolve issues through natural language.

Deploy & Scale Servers

Deploy 10 Rust servers in London, then add more as players arrive. The AI runs the full workflow: find the game, pick a deployment spec and location, create the deployment, poll until ready.

Restart & Reconfigure

Restart a server, change its settings, bump the memory allocation, or install and update mods. The AI uses the platform API and per-container access to make the change.

Monitor & Troubleshoot

"Which of my servers are offline?" "How busy has this server been this week?" The AI checks live server status, player counts and uptime history, and reports back with specifics.

See It in Action

Two example conversations, one from a workspace account and one from a customer account. Each session only has the tools for the account that connected.

BuildWorkspace account: Generate Integration Code
You
Write a Node.js script that deploys Minecraft servers across all EU locations and waits for them to be ready
→ get_guide({topic: "workflows"})
→ list_games()
→ list_game_deployment_specs({gameId: 1})
→ list_locations()
AI
Here's a complete script. It fetches EU locations, creates a deployment in each one with your Minecraft deployment spec, then polls each deployment's /jobs until all servers report ready. Includes error handling and retry logic.
ManageCustomer account: Monitor & Take Action
You
Check my servers in Europe and restart any that aren't running
→ list_clusters()
→ get_server_status({id: 421})
→ restart_server({id: 421})
→ get_server_status({id: 421})
AI
Found 12 servers in your European clusters. Server #421 (Rust, London) had stopped. I've restarted it and it's running again. The other 11 were already running.

MCP Tools & Resources

The tools a session gets depend on the account that connects. A workspace member only gets the tools their permissions allow. Each tool has a typed schema so the AI knows what parameters to pass. Some examples from each set:

Workspace accounts

create_deploymentDeploy game servers for one of your games at a data centerAction
scale_deploymentAdd servers to a deployment or remove themAction
list_usersList users registered to your workspaceAction
create_user_clusterGive a user a cluster (a memory allocation at a data center)Action

Customer accounts

list_clustersList your clusters and the game servers in themAction
create_serverCreate a game server in one of your clustersAction
restart_serverRestart a game server by IDAction
install_modInstall a mod, modpack or collection on a game serverAction

Every account

get_guideStep-by-step workflow guides for multi-step tasksGuide
// Workspace account: AI calls action tools directly
list_games()
→ Minecraft (gameId: 1), Valheim (gameId: 7), ...
list_game_deployment_specs({ gameId: 1 })
→ Survival (specId: 10), ...
create_deployment({ gameId: 1, deploymentSpecId: 10, locationId: 3, count: 2 })
→ Deployment queued. brandDeploymentId: 100, jobId: 456
list_deployment_jobs({ id: 100 })
→ status: completed, 2/2 servers ready
// Workspace account: AI fetches the workflow guide
get_guide({ topic: "workflows" })
// Returns step-by-step:
1. list_games: find your game ID
2. list_game_deployment_specs: pick a deployment spec
3. list_locations: choose a data center
4. create_deployment: provision servers
5. list_deployment_jobs: poll until ready
6. get_deployment: verify result

Frequently Asked Questions

What is MCP?

MCP (Model Context Protocol) is an open standard from Anthropic. It gives AI assistants a way to connect to external tools and data sources over HTTP. One protocol works across Cursor, ChatGPT, Claude, and any other compatible client.

What can I build with Pingcore MCP?

Anything that talks to the Pingcore API. The AI sees a typed schema for every tool your account can use, plus step-by-step workflow guides, so it can generate working integration code in any language. Control panels, deployment scripts, monitoring dashboards.

What can I manage with Pingcore MCP?

It depends on the account you connect. Workspace owners and members can run deployments and fleets, manage users and their clusters, edit game configs, and inspect Discovery and matchmaking, limited to the permissions they hold in the workspace. Customers who rent game servers can create and run servers in their own clusters: restart them, change settings and memory, install mods, and edit config files and scheduled tasks. A few operations, such as billing, stay in the web panel.

Is it free?

Yes. The MCP server is included with every Pingcore plan at no extra cost. It's the same API, just accessed through your AI client instead of curl or code.

Does it work with self-hosted AI?

Yes. The MCP server speaks standard HTTP, so any MCP-compatible client works. That includes open-source tools, self-hosted LLM setups, and custom integrations.

Is it secure?

AI clients authenticate via OAuth 2.1 with PKCE. Your credentials are never exposed to the AI model. Access tokens are opaque, expire after an hour, and can only do what your own account is allowed to do.

Ready to Build and Manage with AI?

Connect your AI assistant to PingCore in under a minute. One URL gets you API documentation, workflow guides, and server management.

Get StartedAPI Documentation