Connect a client

Context Engine runs as a local stdio MCP server. The client starts context-engine serve and supplies your API key through CONTEXT_ENGINE_API_KEY.

For Codex, Claude Code, and Cursor, put the key in your shell profile:

export CONTEXT_ENGINE_API_KEY="<your key>"

Keep the key private and do not commit it in a workspace configuration.

Claude Code

Register Context Engine from the command line:

claude mcp add context-engine -- context-engine serve

Claude Code inherits CONTEXT_ENGINE_API_KEY from your environment.

Cursor

Add Context Engine to your Cursor MCP configuration:

{
  "mcpServers": {
    "context-engine": {
      "command": "context-engine",
      "args": ["serve"],
      "env": {
        "CONTEXT_ENGINE_API_KEY": "${env:CONTEXT_ENGINE_API_KEY}"
      }
    }
  }
}

The configuration reads the key from your environment rather than storing it in the file.

Codex

Add Context Engine to your Codex configuration:

[mcp_servers.context-engine]
command = "context-engine"
args = ["serve"]
env_vars = ["CONTEXT_ENGINE_API_KEY"]

Codex forwards the named variable from its environment to Context Engine.

Google Antigravity IDE

In Antigravity IDE, open … → MCP Servers → Manage MCP Servers → View raw config. Add the server to the global ~/.gemini/config/mcp_config.json file:

{
  "mcpServers": {
    "context-engine": {
      "command": "context-engine",
      "args": ["serve"],
      "env": {
        "CONTEXT_ENGINE_API_KEY": "<your key>"
      }
    }
  }
}

Save the file, then refresh MCP servers. Antigravity receives the literal value as CONTEXT_ENGINE_API_KEY, so use the global file to prevent the key from being committed with a workspace.