About Roar AI
Roar AI’s remote MCP server lets teams manage app hosting and an AI model gateway from an AI assistant. It combines deployment, database, domain and spending controls with access to 200+ models through one account.
Key features
- 01Deploy apps from GitHub or git repositories
- 02Inspect logs and manage deployments
- 03Configure Postgres, storage, domains and jobs
- 04Call 200+ AI models and generate images
- 05Manage API keys, budgets and usage
Use cases
- 01Deploy a repository and retrieve its live URL
- 02Investigate a failed deployment using logs
- 03Connect a Postgres database to an app
- 04Compare model responses and review AI spending
MCP support
The hosted remote MCP server exposes tools for managing Roar AI models, apps, databases, domains, keys and team projects. Clients connect at https://cloud.roar-ai.com/mcp and authorize through OAuth.
Configuration
Remote servers need no local install. Pick your client, then copy the config or install in one click.
https://cloud.roar-ai.com/mcpclaude mcp add --transport http com-roar-ai https://cloud.roar-ai.com/mcp{"mcpServers":{"com-roar-ai":{"type":"http","url":"https://cloud.roar-ai.com/mcp"}}}Open in Cursor{"servers":{"com-roar-ai":{"type":"http","url":"https://cloud.roar-ai.com/mcp"}}}Open in VS Code{"mcpServers":{"com-roar-ai":{"type":"http","url":"https://cloud.roar-ai.com/mcp"}}}{"mcpServers":{"com-roar-ai":{"serverUrl":"https://cloud.roar-ai.com/mcp"}}}Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.
Tools
Details
From the project README
The Roar AI MCP server: every AI model and app hosting, run from your AI assistant
Website · MCP docs · Models · Pricing · Sign in
Roar AI is an AI gateway and an app hosting platform for teams. This repository is the home of its remote MCP server, the one address that lets Claude, ChatGPT, Codex, Cursor, VS Code and any other MCP client work in your Roar AI account for you:
https://cloud.roar-ai.com/mcp
Connect once and sign in with your Roar AI account. There is nothing to install, no API key to paste, and no server to run.
What you can ask
"Deploy my GitHub repo and give me the live link."
"My last deploy failed. Find out why and fix it."
"Add a Postgres database to my app and wire it up."
"Point app.example.com at my app."
"Create an API key for my app with a $50 monthly limit."
"What did we spend on AI this week, and on which models?"
"Ask a different model to review this function, and compare the two answers."
Connect it
New to Roar AI? You can create an account during the connect step itself. Your assistant opens a Roar AI sign-in page, you enter your email and a 6-digit code, and you're back in the conversation.
Claude (claude.ai and the desktop app)
Easiest: open Roar AI in Claude's connector directory and click Connect.
Or add it yourself: go to Settings → Connectors → Add custom connector and paste:
https://cloud.roar-ai.com/mcp
Claude Code
Install it as a plugin from this repository:
claude plugin marketplace add Roar-AI-Labs/roar-ai-mcp
claude plugin install roar-ai@roar-ai
Or add the server directly:
claude mcp add --transport http roar https://cloud.roar-ai.com/mcp
The Claude desktop app does not include the claude command. Claude Code is installed separately.
ChatGPT
Go to Settings → Connectors → Advanced, turn on Developer mode, then press Create. Name it Roar AI, set authentication to OAuth, and paste https://cloud.roar-ai.com/mcp. ChatGPT then opens the Roar AI sign-in page for you to approve.
A one-click listing in ChatGPT's plugin directory is on its way. Its package lives in chatgpt-plugin/.
Codex (CLI, IDE extension and the ChatGPT desktop app)
codex mcp add roar --url https://cloud.roar-ai.com/mcp && codex mcp login roar
Or add it to ~/.codex/config.toml, then run codex mcp login roar:
[mcp_servers.roar]
url = "https://cloud.roar-ai.com/mcp"
Cursor
Add to Cursor in one click, or add this to ~/.cursor/mcp.json (or .cursor/mcp.json for one project):
{
"mcpServers": {
"roar": { "url": "https://cloud.roar-ai.com/mcp" }
}
}
VS Code / GitHub Copilot
Add to VS Code in one click, or run MCP: Open User Configuration (or create .vscode/mcp.json) and add:
{
"servers": {
"roar": { "type": "http", "url": "https://cloud.roar-ai.com/mcp" }
}
}
VS Code names this block servers. Every other tool names it mcpServers, and pasting the wrong one fails without an error.
OpenCode
Add this to opencode.json:
{
"mcp": {
"roar": { "type": "remote", "url": "https://cloud.roar-ai.com/mcp", "enabled": true }
}
}
Anything else (Windsurf, Cline, Continue, JetBrains, Zed and more)
Paste https://cloud.roar-ai.com/mcp wherever the tool asks for a remote MCP server. If it uses a config file, this is the standard shape:
{
"mcpServers": {
"roar": { "url": "https://cloud.roar-ai.com/mcp" }
}
}
What it can do
The server exposes around 90 tools. Your assistant picks the right one from what you ask.
| Area | For example |
|---|---|
| AI models | List the 200+ models you can use, send a message to any of them, generate an image. Calls are billed to your Roar AI account under a monthly cap you set. |
| API keys and spend | Create, limit, rotate and revoke API keys, check usage and budgets, and look up a single request when you need to debug it. |
| App hosting | Deploy from GitHub or any git repo, watch a deploy until it finishes, read build and runtime logs, fix a failed deploy, roll back, stop and start. |
| Databases and storage | Add Postgres to an app, check its health, read its schema, run a query, and add file storage. |
| Domains | Add a custom domain or a free one, check its DNS, and choose the primary address. |
| Config and jobs | Set environment variables, schedule cron jobs and run them, and manage webhooks. |
| Projects and team | Organise apps into projects, and add or remove people. |
Security and permissions
- You sign in with OAuth. Your assistant never sees a password or a long-lived key.
- It can only do what you can. Every action goes through the same permission checks as the Roar AI console. A teammate with read-only access gets a read-only assistant.
- Sensitive reads are opt-in. Reading the rows inside your app's database, or prompts stored in your request logs, needs a permission you grant separately.
- One company at a time. If you belong to more than one, you choose which one your assistant works in, and you can switch by asking it.
- AI spend has a cap. The model calls your assistant makes go through their own key with a monthly limit, which you can pause or change in the console.
See Security at Roar AI for how we handle data, residency and compliance.
About Roar AI
Roar AI gives a team two things behind one account:
- An OpenAI-compatible AI gateway. One API key reaches 200+ frontier and open-source models, with fixed published prices, spending limits per key and per project, and one bill. Read the API docs.
- App hosting built for people who vibe-code. Pick a repo and get a live URL, with a database, a domain and logs, without touching a server. Apps run in the region you choose, including in-country hosting for teams that need their data to stay local.
Start at roar-ai.com, or read the quickstart.
What's in this repository
| Path | What it is |
|---|---|
.claude-plugin/ and plugins/roar-ai/ | A Claude Code plugin marketplace with one plugin, the Roar AI MCP server |
chatgpt-plugin/ | The package for OpenAI's ChatGPT and Codex plugin directory |
server.json | The listing for the official MCP Registry |
docs/publishing.md | How each listing is updated (for maintainers) |
The server itself is hosted by Roar AI, so there is no server code here, only the files that tell each assistant where to find it.
Support
- Something not working? Open an issue or contact us.
- Read the MCP guide and the troubleshooting page.
License
The configuration files in this repository are MIT licensed. Using the Roar AI service is covered by the Roar AI terms and privacy policy.
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.