§ 00Integrations

Connect Server4Agent to your AI assistant, editor, or CLI

Server4Agent speaks the Model Context Protocol, so any AI assistant or agent framework that supports MCP can connect to it: AgentX, Claude, ChatGPT, Muse, Grok Bot, Cursor, VS Code, Codex CLI, GitHub Copilot, Grok Build, Gemini CLI, and anything else that speaks the protocol. Drop one config block into your client and your agent gets a persistent server in a single connection.

Chat and desktop

Editors

Command line

Anywhere else

Claude

One-click connect: add the custom connector, then sign in. No API key to copy.

In-app connector settings

Claude connector menu showing Server4Agent enabled — What it looks like once it is connected: Server4Agent under Connectors, switched on.

https://mcp.server4agent.com

ChatGPT

Add Server4Agent as a custom MCP server, then sign in. No API key to copy.

ChatGPT's Plugins page

ChatGPT composer menu showing Server4Agent available — What it looks like once it is connected: Server4Agent in the composer's plus menu.

https://mcp.server4agent.com

Muse

Ask Muse to add Server4Agent as a custom connector, then sign in. No API key to copy.

A Muse chat

Create a Custom Connector for a new remote MCP server, then connect to it:
Name: Server4Agent
Transport: remote streamable HTTP
URL: https://mcp.server4agent.com
Auth: OAuth

Grok Bot

Add Server4Agent as a custom connector on grok.com, then sign in. Grok Bot can use it from then on.

grok.com/connectors

https://mcp.server4agent.com

AgentX

Connect Server4Agent as a tool in your AgentX workspace.

Workspace tool settings

MCP server URL
https://mcp.server4agent.com

HTTP header
Authorization: Bearer sk_live_YOUR_API_KEY

Cursor

One block for both surfaces: the editor and the cursor-agent CLI read the same config.

~/.cursor/mcp.json

{
  "mcpServers": {
    "server4agent": {
      "url": "https://mcp.server4agent.com",
      "headers": {
        "Authorization": "Bearer sk_live_YOUR_API_KEY"
      }
    }
  }
}

VS Code

.vscode/mcp.json

{
  "servers": {
    "server4agent": {
      "type": "http",
      "url": "https://mcp.server4agent.com",
      "headers": {
        "Authorization": "Bearer sk_live_YOUR_API_KEY"
      }
    }
  }
}

Codex CLI

The one client here that configures MCP in TOML, not JSON.

~/.codex/config.toml

# ~/.codex/config.toml
[mcp_servers.server4agent]
url = "https://mcp.server4agent.com"
bearer_token_env_var = "SERVER4AGENT_API_KEY"

GitHub Copilot

For the Copilot CLI. Using Copilot inside your editor? The VS Code tab is the one you want.

~/.copilot/mcp-config.json

{
  "mcpServers": {
    "server4agent": {
      "type": "http",
      "url": "https://mcp.server4agent.com",
      "headers": {
        "Authorization": "Bearer ${SERVER4AGENT_API_KEY}"
      }
    }
  }
}

Grok Build

One command, because this client reads config from more places than any other.

grok mcp add, or .grok/mcp.json

grok mcp add --transport http server4agent https://mcp.server4agent.com \
  --header "Authorization: Bearer ${SERVER4AGENT_API_KEY}"

Gemini CLI

One command adds Server4Agent to your Gemini CLI settings.

~/.gemini/settings.json

gemini mcp add --scope user --transport http server4agent https://mcp.server4agent.com --header "Authorization: Bearer sk_live_YOUR_API_KEY"

Agent Skill

Install Server4Agent as a reusable skill: invoke with /server4agent in any compatible agent.

~/.claude/skills/ or .claude/skills/

# Personal install: available across all your projects
mkdir -p ~/.claude/skills/server4agent
curl -o ~/.claude/skills/server4agent/SKILL.md \
  https://www.server4agent.com/skill.md

# Project install: available to everyone in the repo
mkdir -p .claude/skills/server4agent
curl -o .claude/skills/server4agent/SKILL.md \
  https://www.server4agent.com/skill.md

Any MCP host

Works with any AI assistant or agent framework that speaks MCP.

Your client's MCP config

{
  "url": "https://mcp.server4agent.com",
  "headers": {
    "Authorization": "Bearer sk_live_YOUR_API_KEY"
  }
}

The endpoint is the same everywhere: https://mcp.server4agent.com. Pick your client above, or use the raw JSON with any MCP-compatible host.

Deploying from ChatGPT? Follow the step-by-step guide.

§ 01Three steps

From config block to a live URL

The flow is the same in every host; only where you paste the config changes.

Step 01

Create an API key

Generate a scoped key in the dashboard. It authenticates your agent to Server4Agent and is revocable at any time.

Step 02

Add the MCP server

Paste the config block into your client's MCP settings and swap in your key. One endpoint, no glue code.

Step 03

Let your agent build

Your agent now has Server4Agent's tools natively. It provisions a server, writes the code, runs it, and ships a live URL.

One connection, the full toolset

MCP is host-agnostic on purpose. Whatever client your agent runs in, the same Server4Agent tools show up, so you wire it once and the agent does the rest.

What you get

▸Provision and control persistent servers as native tools
▸A single prompt tool the on-server agent plans and executes
▸Shell, file, build, and deploy primitives the agent can reach for
▸Scoped, individually revocable API keys per client

Ready to wire up your agent?

Create a key, paste the config, and your agent has a server it can build on. The full schema and tool reference live in the MCP docs.

Get early access →MCP docs →