Your AI builds, deploys, and runs full-stack apps on a hosted workspace created at first sign-in.
MCP Server: co.obto/obto
The MCP server co.obto/obto provides an integration point for building, deploying, and running full-stack applications. It uses a hosted workspace that is created the first time a user signs in, and then supports subsequent development and execution workflows within that environment.
🛠️ Key Features
Builds full-stack apps
Deploys applications
Runs applications
Uses a hosted workspace created at first sign-in
🚀 Use Cases
Creating full-stack projects from within an MCP-connected workflow
Deploying completed applications to an environment managed by the service
Running applications after deployment in the hosted workspace
⚡ Developer Benefits
Centralizes build/deploy/run actions for full-stack apps
Automatically provisions a workspace on first sign-in
⚠️ Limitations
No additional capabilities (e.g., specific tools, tool count, or APIs) are described in the provided server data.
The first MCP-native platform where AI agents deploy production applications.
Describe what you want. The AI builds it and deploys it. You get a live URL.
OBTO is a full-stack deployment platform built for the age of AI agents. Connect it to Claude, GPT, Codex, or any MCP-compatible model. Describe what you want in plain language. The AI writes the code, provisions the database, and deploys your application to a live URL — all in the same conversation.
No code editor. No terminal. No hosting setup. Full-stack. Not a wrapper over Vercel and Supabase — real backend, real infrastructure, one platform.
OBTO has been running in production since 2019.
Quick Start
1. Sign Up
Go to obto.co and click Sign in. Authenticate with Google. Your OBTO instance is created automatically.
2. Copy Your MCP Endpoint
After signing in, you'll see your personal MCP server URL on the landing page. Copy it.
3. Teach Your AI Client — plugin or skill file
Install the plugin (recommended where supported). This repository is itself the plugin marketplace — one URL, nothing else to configure:
Installs the OBTO MCP server connection plus six task skills (build loop, deploys & large files, media upload, memory, MCP extension, troubleshooting). Details: plugins/obto.
OpenAI Codex — clone this repository, then codex plugin marketplace add "/path/to/platform" and install OBTO. The Codex package is plugins/obto-codex.
Use the skill files directly (every other harness). The skills are plain-markdown SKILL.md folders — portable to any agent harness that reads skills or instruction files:
Client
What to use
Claude Desktop
Add SKILL.md to your project's knowledge or attach as a file
Cursor
Place SKILL.md in your workspace root or .cursor/skills/
Add your MCP server URL to your AI client's MCP configuration:
Claude Desktop — Go to Settings → MCP Servers → Add your OBTO endpoint URL. Claude will redirect you to authenticate via OAuth. Approve it.
Claude Web — In a project, add the MCP server URL. Authorize when prompted.
Cursor / VS Code — Add the MCP server configuration to your settings.
5. Build
Type what you want. The AI handles the rest.
code
"Build me a landing page for my coffee shop with a menu, gallery, and contact form."
Your app will be live at yourapp.obto.co.
What Does the Skill File Do?
The SKILL.md file is a set of instructions that teaches AI models how to build and deploy applications on OBTO. It covers:
Platform architecture — How OBTO stores code as database artifacts (no filesystem)
Deployment order — Backend first, frontend last
Collection types — Pages, JavaScript modules, stylesheets, routes, server scripts
The xe. pattern — How backend services call each other
MCP tool routing — Which tool to use for each operation
Full-stack examples — Complete contact form deployment walkthrough
Without this skill file, the AI won't know OBTO's conventions and will produce code that doesn't deploy correctly. With it, the AI becomes an autonomous developer on your OBTO instance.
Where your client supports plugins (Claude Code, Claude Cowork, Codex), prefer the plugin — the same doctrine split into six auto-loading task skills, with the MCP connection preconfigured.