Most developers using Claude Code or Cursor work from their local machine. The AI edits local files, runs local commands, and operates within the boundaries of whatever is installed on the developer’s computer. Zo Computer changes that infrastructure equation.
Your persistent cloud Linux server becomes the backend for your AI tools. Your IDE gets access to cloud files, connected apps, and a full tool suite it cannot reach locally. The development environment and the deployment environment collapse into one.
This guide covers the full developer picture: the MCP server, GitHub, documentation, code execution, and how to use Zo as a persistent cloud development environment.
Why developers use Zo
Three things make Zo different from a local dev environment or a standalone AI tool.
- Persistent compute. Your server stays on with the Basic plan. Scripts run, services stay live, and context persists across sessions without manual restart.
- Cloud files with AI access. Your AI has native read/write access to your cloud filesystem. Files you store in Zo are available to any AI client you connect via MCP.
- 50+ built-in tools. Web search, shell execution, file operations, Gmail, Notion, Airtable, and more, all accessible from within your AI client without additional integration work.
Together, these make Zo a unified development environment where your AI and your cloud infrastructure share the same context.
The Zo MCP Server
The MCP Server is the core developer feature. MCP (Model Context Protocol) is an open standard that lets AI clients connect to external tools and data sources. The Zo MCP Server implements this protocol, so any MCP-compatible AI client can talk to your Zo.
What the MCP server gives your AI client access to:
- Your Zo cloud files and filesystem
- Data from connected apps (Gmail, Notion, Airtable, Linear, Dropbox)
- Zo’s full tool suite including web search, shell commands, and code execution
- Any skills or agents you have configured in your Zo
Supported AI clients
| AI tool | MCP connection method | What Zo gives it access to |
|---|---|---|
| Claude Code | Zo MCP Server config | Cloud files, Gmail, Notion, Airtable, 50+ tools, shell |
| Cursor | Zo MCP Server config | Cloud files, connected apps, Zo tool suite |
| Gemini CLI | Zo MCP Server config | Cloud files, connected apps, Zo tool suite |
| Codex | Zo MCP Server config | Cloud files, connected apps, Zo tool suite |
All four clients use the same Zo MCP Server. The configuration differs slightly per client but the capabilities are the same.
How to connect Claude Code to Zo via MCP
The full step-by-step setup is documented at zocomputer.mintlify.app. The high-level steps are:
- Install the Zo MCP Server following the docs for your platform.
- Add the Zo MCP Server to your Claude Code MCP configuration file.
- Authenticate the connection with your Zo credentials.
- Restart Claude Code and verify the connection is active.
Once connected, Claude Code can read and write files in your Zo cloud, run shell commands on your Zo server, and call any of Zo’s built-in tools, all from within your normal Claude Code workflow.
For more on getting started with Zo generally, see Getting started with Zo Computer.
For a broader view of what Claude Code can do, see What is Claude Code or take the Claude Code course.
GitHub: open-source components
Zo’s GitHub repository is at github.com/zocomputer/Zo. The repository description is: “Create powerful personal software on your intelligent cloud computer.”
What the GitHub repository contains:
The repo includes open-source components of the Zo platform. Developers building on top of Zo, contributing to the project, or wanting to understand how specific features work can reference it directly.
Zo launched in June 2026 and is described as the original inspiration for OpenClaw, the open-source self-hosted alternative. The Zo GitHub repo reflects that heritage — it is built with transparency about how the platform works.
Documentation
The full developer documentation lives at zocomputer.mintlify.app. It covers:
- MCP server installation for each supported client
- File system API and file operations
- Task scheduling configuration
- Integration setup for Gmail, Notion, Airtable, and others
- Shell access and code execution
- Skills registry and agent configuration
- Troubleshooting guides for common setup issues
This is the reference you should go to for specific syntax, config file formats, and step-by-step setup flows. The guide you are reading now gives the conceptual overview. The docs give the exact commands.
Bring your own API keys
All Zo plans support bringing your own API keys. This matters for two reasons.
First, it gives you cost control. AI usage on Zo is charged at cost, the same rates Zo pays providers. You do not pay a markup. If you bring your own keys from Anthropic, OpenAI, or another provider, you use your existing credits directly.
Second, it gives you model flexibility. You are not locked into whatever models Zo has selected as defaults. You can use the model that fits your task and budget.
The result: Bringing your own keys is especially useful for developers running high-volume automated tasks, where AI costs can accumulate quickly.
Native code execution
Zo supports sandboxed Python and JavaScript execution natively on your cloud server. This is different from generating code in a chat window and copying it elsewhere.
What native execution means:
You write a script in conversation with Zo (or upload one to your cloud filesystem), and Zo runs it on your server. The output comes back in the same session. No local environment required. No dependency conflicts.
- Python. Useful for data analysis, scraping, automation scripts, and report generation.
- JavaScript. Useful for web-adjacent tasks, API calls, and data transformation.
Both execution environments are sandboxed: they run in isolation and cannot affect the host system. For a discussion of the security model, see Zo Computer security and privacy.
Shell access via MCP
When you connect Claude Code or Cursor to your Zo via MCP, your AI client can run shell commands on your Zo server. This is not running commands locally. The commands execute in your cloud environment.
Practical implications:
- Install packages on your cloud server from within your IDE.
- Run build scripts on a remote, always-on environment.
- Trigger cron jobs or automation scripts through your AI client.
- Access cloud-stored files via shell without downloading them first.
Shell access via MCP collapses the distinction between “asking your AI to do something” and “running a command on a server.” For workflows that require both, this is the most direct integration available.
Zo as a persistent cloud development environment
Local development environments are stateful but local. Cloud environments are persistent but often disconnected from your AI tools. Zo connects both.
Your Zo server is always on (Basic plan). Files you store there persist across sessions. AI clients connected via MCP have continuous access to those files and can operate on them at any time, including via scheduled tasks that run without you.
What this unlocks for developers:
- Run a test suite on a schedule, get results via email.
- Keep a development server live at a permanent URL for stakeholder demos.
- Store project context (notes, specs, reference files) in your cloud and make it available to every AI tool you use.
- Run long-running scripts without keeping your laptop awake.
The Basic plan at $18/month is the practical entry point for developers who want the always-on behavior. The free plan sleeps when inactive, which limits scheduled and persistent use cases.
Common questions on Zo Computer for developers
”Does the MCP server work with both Claude Code and Cursor simultaneously?”
Yes. The Zo MCP Server can be connected to multiple AI clients. Each client authenticates separately and gets access to the same Zo environment. You can have Claude Code and Cursor both connected to the same Zo cloud simultaneously.
”Can I use the MCP server on the free plan?”
Yes. The Zo MCP Server is available on all plans, including free. The free plan limitation is server sleep when inactive. If you want MCP access to a server that is always on and ready to respond, the Basic plan is required.
”Where do I find the MCP configuration syntax for Claude Code specifically?”
The documentation at zocomputer.mintlify.app has client-specific configuration guides. Claude Code’s MCP config uses a JSON file that you edit to add the Zo server endpoint and credentials. The exact syntax is in the docs under the MCP setup section.
”Is Zo safer than giving Claude Code access to my local machine?”
Yes, for one specific reason: Zo runs in an isolated cloud environment, not your local file system. If an AI client makes an unintended change or runs an unintended command, the blast radius is your Zo cloud environment, not your local machine. Your local files, applications, and system are unaffected. Note: For more on this, see Zo Computer security and privacy.
Get your development environment connected
The fastest path is installing the MCP server and connecting Claude Code or Cursor today. Full docs at zocomputer.mintlify.app. GitHub at github.com/zocomputer/Zo.
The developers getting the most from Zo are not just using it as a chat interface. They have connected their IDE via MCP and built their cloud environment into their standard workflow. That shift — from AI as advisor to AI as infrastructure — is where the time savings compound.
Path one: install the MCP server this week. Go to zocomputer.mintlify.app and follow the setup guide for your AI client. Connect Claude Code or Cursor to your Zo. Run one command through the MCP connection and see what it gives you access to.
Path two: bring in a partner. Phos AI Labs helps development teams set up AI infrastructure, including Zo MCP integration, Claude Code workflows, and persistent cloud environments, that fits how their team actually works. Thirty minutes, no deck. Start here.
Related articles