Getting Started

Imagine. Build. Ship.

If you can imagine it, OBTO can build it. Connect your AI client and deploy your first application through conversation.

⚡ Fastest path — Claude Code

One install wires up everything. The OBTO plugin bundles the MCP connection, all six skills, and the /obto:start command — so you can skip Steps 2–4.

Claude Code
/plugin marketplace add obto-inc/platform
/plugin install obto@obto

Using Cursor, VS Code, Claude.ai, ChatGPT, or another MCP client? Follow the four steps below.

1

Connect with OBTO MCP

OBTO uses one universal endpoint — the same URL for everyone. Nothing to generate, no API key. Copy it; you'll paste it into your AI client in Step 2.

Universal MCP Endpoint
https://app.obto.co/ms/mcp

💡 You can sign up at obto.co with Google first — or skip it. The first time your client connects to this endpoint, OAuth signs you in and provisions your workspace automatically.

2

Add It to Your AI Client

Pick your client below — every config uses the one universal endpoint https://app.obto.co/ms/mcp from Step 1. Copy & paste as-is.

Claude.ai — Web Interface

The simplest option. Native remote MCP via Connectors.

1

Settings → Connectors (claude.ai/settings/connectors)

2

Click "Add custom connector"

3

Paste your OBTO MCP endpoint URL

4

Click "Add" — transport auto-detected

5

In chat: "+""Connectors" to toggle on

Plan Note

Free: 1 connector. Pro/Max/Team/Enterprise: unlimited. Team/Enterprise orgs add via Org settings → Connectors first.

Claude Desktop App

Uses Connectors UI for remote servers — not the config file.

Recommended: Connectors UI

1

"+" at bottom of chat → "Connectors""Manage Connectors"

2

Add custom connector with your endpoint URL

3

Syncs across Claude.ai, Desktop, and Mobile

Alternative: Config file with mcp-remote

Requires Node.js. Bridges stdio to HTTP.

macOS~/Library/Application Support/Claude/claude_desktop_config.json Windows%APPDATA%\Claude\claude_desktop_config.json Linux~/.config/claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "obto": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://app.obto.co/ms/mcp"
      ]
    }
  }
}
Common Pitfall

The config file does not support a direct "url" field. {"url":"https://..."} silently fails. Use Connectors UI or mcp-remote. Fully quit & restart after changes.

Claude Code (CLI)

Native HTTP support — simplest CLI setup.

One-command setup

Terminal
claude mcp add --transport http obto-mcp https://app.obto.co/ms/mcp

With scope

Terminal
# User scope (all projects)
claude mcp add --transport http --scope user obto-mcp https://app.obto.co/ms/mcp

# Project scope (team via .mcp.json)
claude mcp add --transport http --scope project obto-mcp https://app.obto.co/ms/mcp
Flag Order

All flags must come before the server name. claude mcp add obto-mcp --transport http fails.

Config file

~/.claude.json (user) or .mcp.json (project root):

.mcp.json
{
  "mcpServers": {
    "obto-mcp": {
      "type": "http",
      "url": "https://app.obto.co/ms/mcp"
    }
  }
}

Useful: claude mcp list · claude mcp get obto-mcp · /mcp · claude --mcp-debug

VS Code with GitHub Copilot (Agent Mode)

Built-in since VS Code 1.99+. Requires Copilot extension + chat.agent.enabled: true.

Option A — Workspace (recommended)

Create .vscode/mcp.json:

.vscode/mcp.json
{
  "servers": {
    "obto": {
      "type": "http",
      "url": "https://app.obto.co/ms/mcp"
    }
  }
}

Option B — User settings.json

settings.json
{
  "mcp": {
    "servers": {
      "obto": {
        "type": "http",
        "url": "https://app.obto.co/ms/mcp"
      }
    }
  }
}
Critical

VS Code uses "servers"not "mcpServers". #1 config error. Verify: Copilot Chat → Agent mode → tools icon.

VS Code with Continue.dev

YAML config. Tools only in Agent mode.

macOS/Linux~/.continue/config.yaml Windows%USERPROFILE%\.continue\config.yaml
config.yaml
mcpServers:
  - name: OBTO
    type: streamable-http
    url: https://app.obto.co/ms/mcp
Tip

If streamable-http fails, try sse. Continue uses kebab-case: sse, stdio, streamable-http.

VS Code with Cline

Access: MCP Servers icon → Configure → Configure MCP Servers

macOS~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json Windows%APPDATA%\Code\User\globalStorage\saoudrizwan.claude-dev\settings\cline_mcp_settings.json
cline_mcp_settings.json
{
  "mcpServers": {
    "obto": {
      "url": "https://app.obto.co/ms/mcp",
      "type": "sse",
      "disabled": false,
      "timeout": 60
    }
  }
}
Tip

Start with "sse" — more reliable than "streamableHttp" (Cline uses camelCase). Green dot = connected.

VS Code with Roo Code

Access: 🔌 MCP icon → Edit Global/Project MCP

macOS~/Library/Application Support/Code/User/globalStorage/rooveterinaryinc.roo-cline/settings/mcp_settings.json Windows%APPDATA%\Code\User\globalStorage\rooveterinaryinc.roo-cline\settings\mcp_settings.json Project.roo/mcp.json (overrides global)
mcp_settings.json
{
  "mcpServers": {
    "obto": {
      "type": "sse",
      "url": "https://app.obto.co/ms/mcp",
      "disabled": false,
      "timeout": 60
    }
  }
}
Required

type is required — omitting it errors. Kebab-case: "sse", "streamable-http", "stdio".

Cursor IDE

Native remote MCP. Simplest IDE config.

Project.cursor/mcp.json Global~/.cursor/mcp.json UIPreferences → Cursor Settings → Tools & MCP
.cursor/mcp.json
{
  "mcpServers": {
    "obto": {
      "url": "https://app.obto.co/ms/mcp"
    }
  }
}

Auto-detects transport. No "type" needed. Green indicator in Settings → Tools & MCP.

Tool Limit

Cursor caps at ~40 tools across all servers. Toggle off/on or restart if tools don't appear.

Windsurf IDE (Codeium)

Own config path, different key name for URL.

macOS/Linux~/.codeium/windsurf/mcp_config.json Windows%USERPROFILE%\.codeium\windsurf\mcp_config.json UISettings → Cascade → Manage MCPs
mcp_config.json
{
  "mcpServers": {
    "obto": {
      "serverUrl": "https://app.obto.co/ms/mcp"
    }
  }
}
Windsurf Key

"serverUrl" not "url". Most common Windsurf MCP error.

JetBrains IDEs

All IntelliJ-based IDEs since 2025.1.

1

Settings → Tools → AI Assistant → Model Context Protocol (MCP)

2

+ AddHTTP type

3

Paste config, OK → Apply

JetBrains MCP
{
  "mcpServers": {
    "obto": {
      "url": "https://app.obto.co/ms/mcp"
    }
  }
}
Tip

"Import from Claude" button imports existing Claude Desktop configs.

Zed Editor

Stdio only — needs mcp-remote (Node.js required).

Zed settings.json
{
  "context_servers": {
    "obto": {
      "source": "custom",
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://app.obto.co/ms/mcp"],
      "env": {}
    }
  }
}
Zed Quirk

"source": "custom" is required or Zed silently skips the server.

Amazon Q Developer

Native HTTP. Config: ~/.aws/amazonq/default.json or .amazonq/default.json:

default.json
{
  "mcpServers": {
    "obto": {
      "type": "http",
      "url": "https://app.obto.co/ms/mcp"
    }
  }
}

Universal Fallback (Any stdio-only Client)

mcp-remote bridges stdio to HTTP. Wrap in whatever root key your client expects:

Generic bridge
{
  "command": "npx",
  "args": ["-y", "mcp-remote", "https://app.obto.co/ms/mcp"]
}
3

Install the OBTO Skill

The OBTO Skill (obto-developer) teaches your AI model how to build and deploy on OBTO. Install it into your coding agent with a single command — no manual file placement:

Terminal
npx skills add obto-inc/platform

Your agent is auto-detected — Claude Code, Codex, Cursor, Windsurf, and 70+ others — and the skill lands in the right place. Handy variants:

Terminal
# Preview the skill before installing
npx skills add obto-inc/platform --list

# Install into a specific agent
npx skills add obto-inc/platform -a claude-code
Powered by the open skills CLI

Installation runs on vercel-labs/skills, the open agent-skills installer — it pulls the skill straight from github.com/obto-inc/platform. Requires Node.js (bundled with npx).

Chat clients (Claude.ai / Desktop)

No skills directory? Grab SKILL.md from GitHub and attach it to your conversation or project knowledge.

4

Build

You're connected. Select the obto-developer skill and then just describe what you want:

"Build me a landing page for my coffee shop called Daily Grind with a hero section, menu with prices, photo gallery, and a contact form that saves submissions."

The AI reads the OBTO Skill, writes the backend logic, creates the frontend, deploys everything in the correct order, and your app goes live at yourapp.obto.co.

Ready to build?

Sign in and get your MCP endpoint in under a minute.

Get Started →