AI Assistant Integration

Connect Source Parts to OpenAI Codex

Expose the Source Parts MCP server as a trusted tool inside OpenAI Codex so engineering copilots can reason over validated component data.

Register the remote MCP endpoint

Codex accepts remote MCP servers via the `mcp_server` tool definition. Point to the Source Parts managed endpoint or your own proxy depending on compliance requirements.

For managed hosting keep `url` pointed at `https://api.source.parts/mcp`.
When self-hosting, place the MCP server behind your preferred API gateway to enforce corporate network policies.

{
  "tools": [
    {
      "mcp_server": {
        "name": "sourceparts",
        "url": "https://api.source.parts/mcp",
        "api_key": "$SOURCEPARTS_API_KEY"
      }
    }
  ]
}

Provide task-specific prompts

Codex performs best when you hint at which tools to use. Encourage prompts like “use `inventory.search` to gather options before answering.”

Bundle canonical prompt templates with your IDE extension so developers do not have to remember individual tool payloads.
Log Codex tool invocations through your existing observability stack to catch auth misconfigurations early.