MCP Setup Guide

This guide walks you through adding Apart Intelligence to every major AI tool that supports MCP. Each section includes the exact config, file path, and our recommendation for the best way to use Apart Intelligence with that tool.

Note

Every configuration below requires your API URL and API key. Get your key by running ai login with the CLI or from the Apart Intelligence dashboard. Replace apart_your-api-key with your actual key.

Quick Recommendations

ToolBest approach
Claude CodeUse the CLI skills (ai codebase init) for the best experience. Add the MCP as well for full tool access.
Codex CLIUse the CLI alongside the MCP. Terminal tools benefit from having both the CLI for direct commands and the MCP for AI-driven access.
Gemini CLIUse the CLI for direct commands and the MCP for AI-driven access, same as Codex CLI.
Cursor / Windsurf / VS Code / ClineUse the MCP. IDE-based tools get the most value from the MCP since the AI agent can search, create, and link knowledge inline.
Claude DesktopUse the MCP. Ideal for non-coding workflows like capturing decisions, searching knowledge, or reviewing graph health.
JetBrains / ZedUse the MCP. Same reasoning as Cursor — the AI agent benefits from inline tool access.

Claude Code

Claude Code is a terminal-based AI tool from Anthropic. We recommend using both the CLI skills and the MCP for the best experience. The CLI skills give you slash commands like /capture, /document, and /sync. The MCP gives Claude full tool access to search, create, and manage knowledge.

Install CLI skills (recommended)

ai codebase init

Add the MCP server

Use the claude mcp add command. All flags must come before the server name, and use -- to separate the name from the command:

claude mcp add \
  --transport stdio \
  --env APART_API_URL=https://api.intelligence.apart.tech \
  --env APART_API_KEY=apart_your-api-key \
  apart-intelligence -- npx -y @apart-tech/intelligence-mcp

To share the MCP config with your team, add it at project scope instead. Create a .mcp.json file in the project root:

.mcp.json
{
  "mcpServers": {
    "apart-intelligence": {
      "command": "npx",
      "args": ["-y", "@apart-tech/intelligence-mcp"],
      "env": {
        "APART_API_URL": "https://api.intelligence.apart.tech",
        "APART_API_KEY": "apart_your-api-key"
      }
    }
  }
}

Tip

Project-scoped servers in .mcp.json require user approval on first use. Team members will see a one-time security prompt.

Cursor

Cursor supports MCP via project-level and global config files. For a per-project setup, create .cursor/mcp.json in the project root. For global access, use ~/.cursor/mcp.json.

.cursor/mcp.json
{
  "mcpServers": {
    "apart-intelligence": {
      "command": "npx",
      "args": ["-y", "@apart-tech/intelligence-mcp"],
      "env": {
        "APART_API_URL": "https://api.intelligence.apart.tech",
        "APART_API_KEY": "apart_your-api-key"
      }
    }
  }
}

Warning

Cursor has a hard limit of roughly 40 active tools across all MCP servers combined. If you use multiple MCP servers, keep an eye on your total tool count.

VS Code (GitHub Copilot)

VS Code requires version 1.99+ and uses a different config format than most other tools. The top-level key is servers (not mcpServers), and each server needs an explicit type field.

Create .vscode/mcp.json in the project root (commit it to share with your team), or configure via VS Code Settings.

.vscode/mcp.json
{
  "inputs": [
    {
      "type": "promptString",
      "id": "apart-api-key",
      "description": "Apart Intelligence API Key",
      "password": true
    }
  ],
  "servers": {
    "apart-intelligence": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@apart-tech/intelligence-mcp"],
      "env": {
        "APART_API_URL": "https://api.intelligence.apart.tech",
        "APART_API_KEY": "${input:apart-api-key}"
      }
    }
  }
}

Tip

The inputssystem prompts for your API key on first use and stores it securely — no plaintext keys in your repo.

Windsurf

Windsurf stores MCP config globally. Edit the config file at:

  • macOS/Linux: ~/.codeium/windsurf/mcp_config.json
  • Windows: %USERPROFILE%\.codeium\windsurf\mcp_config.json
mcp_config.json
{
  "mcpServers": {
    "apart-intelligence": {
      "command": "npx",
      "args": ["-y", "@apart-tech/intelligence-mcp"],
      "env": {
        "APART_API_URL": "https://api.intelligence.apart.tech",
        "APART_API_KEY": "apart_your-api-key"
      }
    }
  }
}

You can also install MCP servers from the Cascade panel's MCP Marketplace.

Claude Desktop

Claude Desktop is great for non-coding workflows — searching knowledge, capturing decisions, and reviewing graph health. Edit the config file:

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "apart-intelligence": {
      "command": "npx",
      "args": ["-y", "@apart-tech/intelligence-mcp"],
      "env": {
        "APART_API_URL": "https://api.intelligence.apart.tech",
        "APART_API_KEY": "apart_your-api-key"
      }
    }
  }
}

Note

Restart Claude Desktop after editing the config for changes to take effect.

OpenAI Codex CLI

Codex CLI is a terminal-based tool similar to Claude Code. Like Claude Code, we recommend pairing the CLI with the MCP for the best experience. Codex uses TOML (not JSON) for config.

Add to ~/.codex/config.toml (user-level) or .codex/config.toml (project-level):

config.toml
[mcp_servers.apart-intelligence]
command = "npx"
args = ["-y", "@apart-tech/intelligence-mcp"]
env = { "APART_API_URL" = "https://api.intelligence.apart.tech", "APART_API_KEY" = "apart_your-api-key" }

Or add it via the CLI:

codex mcp add apart-intelligence -- npx -y @apart-tech/intelligence-mcp

Gemini CLI

Google's Gemini CLI supports MCP via its settings file. Add to ~/.gemini/settings.json (global) or .gemini/settings.json (project-level):

settings.json
{
  "mcpServers": {
    "apart-intelligence": {
      "command": "npx",
      "args": ["-y", "@apart-tech/intelligence-mcp"],
      "env": {
        "APART_API_URL": "https://api.intelligence.apart.tech",
        "APART_API_KEY": "apart_your-api-key"
      }
    }
  }
}

Cline

Cline is a VS Code extension with its own MCP configuration. The easiest way is via the Cline MCP panel in VS Code, but you can also edit the config file directly:

  • macOS: ~/Library/Application Support/Code/User/globalStorage/saoudrizwan.claude-dev/settings/cline_mcp_settings.json
  • Linux: ~/.config/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": {
    "apart-intelligence": {
      "command": "npx",
      "args": ["-y", "@apart-tech/intelligence-mcp"],
      "env": {
        "APART_API_URL": "https://api.intelligence.apart.tech",
        "APART_API_KEY": "apart_your-api-key"
      }
    }
  }
}

JetBrains IDEs

IntelliJ IDEA, WebStorm, PyCharm, and other JetBrains IDEs (2025.2+) support MCP via Settings > Tools > AI Assistant > Model Context Protocol. For project-level config, create .junie/mcp/mcp.json:

.junie/mcp/mcp.json
{
  "mcpServers": {
    "apart-intelligence": {
      "command": "npx",
      "args": ["-y", "@apart-tech/intelligence-mcp"],
      "env": {
        "APART_API_URL": "https://api.intelligence.apart.tech",
        "APART_API_KEY": "apart_your-api-key"
      }
    }
  }
}

Zed

Zed uses a different config key than other tools. Open Zed > Settings > Open Settings and add:

settings.json
{
  "context_servers": {
    "apart-intelligence": {
      "source": "custom",
      "command": "npx",
      "args": ["-y", "@apart-tech/intelligence-mcp"],
      "env": {
        "APART_API_URL": "https://api.intelligence.apart.tech",
        "APART_API_KEY": "apart_your-api-key"
      }
    }
  }
}

Warning

The "source": "custom" field is required. Without it, Zed silently ignores the server entry.

Windows Note

On Windows, some tools may have trouble running npx directly. If you see "Connection closed" errors, wrap the command:

"command": "cmd",
"args": ["/c", "npx", "-y", "@apart-tech/intelligence-mcp"]

This applies to Claude Desktop, Cursor, and any tool using stdio transport on Windows.

Troubleshooting

ProblemSolution
Server not showing upRestart the tool after editing the config file. Most tools do not hot-reload MCP config.
"Connection closed" on WindowsUse the cmd /c npx wrapper shown above.
Authentication errorsVerify your APART_API_KEY starts with apart_ and is valid. Run ai login to re-authenticate.
Tools missing in CursorCheck your total tool count across all MCP servers. Cursor caps at ~40 tools.
Zed ignores the serverEnsure "source": "custom" is set.