> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.paradex.trade/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.paradex.trade/_mcp/server.

# Remote MCP quickstart

> **Warning**
>
> Remote deployments do **not support trading** at the moment. Only public market data is available. Use a [local setup](/agentic-ai-hub/mcp/quickstart) for full API access including trading.

## How it works

The MCP server supports two transport modes controlled by `MCP_TRANSPORT`:

| Mode            | Value             | Use case                                                         |
| --------------- | ----------------- | ---------------------------------------------------------------- |
| stdio (default) | `stdio`           | Claude Desktop, Claude Code, Cursor — runs as a local subprocess |
| HTTP            | `streamable-http` | ChatGPT Desktop, Claude.ai web, any remote client                |

In HTTP mode the server listens on `MCP_PORT` (default `3000`) and exposes:

* `POST /mcp` — MCP protocol endpoint (what clients connect to)
* `GET /health` — health check, returns `{"status": "ok"}`

For more details, see the [deploy-http guide](https://github.com/tradeparadex/mcp-paradex-py/blob/main/docs/deploy-http.md) in the repository.

---

## Environment variables

| Variable              | Description                                         | Default |
| --------------------- | --------------------------------------------------- | ------- |
| `MCP_TRANSPORT`       | Set to `streamable-http` for remote deployment      | `stdio` |
| `MCP_PORT`            | Port the HTTP server listens on                     | `3000`  |
| `PARADEX_ENVIRONMENT` | Target environment: `prod`, `testnet`, or `nightly` | `prod`  |

Most platforms let you set these as environment variables in their dashboard. If you've cloned the MCP server repository to build and deploy, you can also store them in a `.env` file at the repo root:

```bash
cp .env.template .env
```

Then edit `.env` with your values. The server loads this file automatically on startup.

---

## Choose a platform

| Option                                                | Effort | Cost                | Best for                 |
| ----------------------------------------------------- | ------ | ------------------- | ------------------------ |
| [Railway](/agentic-ai-hub/remote-mcp/railway)         | Low    | \~\$5/mo            | Quickest permanent setup |
| [Render](/agentic-ai-hub/remote-mcp/render)           | Low    | Free tier available | Low-traffic / evaluation |
| [Fly.io](/agentic-ai-hub/remote-mcp/flyio)            | Medium | \~\$2–5/mo          | Global, low-latency      |
| [AWS Lambda](/agentic-ai-hub/remote-mcp/aws-lambda)   | Medium | Pay-per-request     | Scalable, zero idle cost |
| [Docker / VPS](/agentic-ai-hub/remote-mcp/docker-vps) | High   | Varies              | Full control             |

All options produce an endpoint of the form `https://your-domain.example.com/mcp` that you paste into your AI client's MCP settings.

---

## Verify your deployment

Test your endpoint from the terminal (replace the URL with your actual deployment URL):

```bash
curl https://your-endpoint.example.com/health
# Expected: {"status":"ok"}

curl -X POST https://your-endpoint.example.com/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"tools/list","id":1}'
# Expected: JSON list of available tools
```

---

## Connect your AI client

#### Claude.ai (Web)

1. Open [claude.ai](https://claude.ai) → **Settings → Integrations**.
2. Click **Add MCP Server**.
3. Enter your endpoint URL (e.g. `https://your-domain.example.com/mcp`).

#### ChatGPT Desktop

1. Open ChatGPT Desktop → **Settings → Advanced → Developer Mode** (enable if not already).
2. Go to **Settings → Connectors → Create**.
3. Enter your endpoint URL (e.g. `https://your-domain.example.com/mcp`).
4. Click **I trust this provider**.