Gemini

Connect Google's Gemini CLI (or other Gemini-based coding assistants) to AI-ERD via MCP.

---

Add the AI-ERD MCP server

Gemini CLI reads MCP server configuration from ~/.gemini/settings.json:

{
  "mcpServers": {
    "ai-erd": {
      "httpUrl": "https://ai-erd.com/mcp"
    }
  }
}

Restart Gemini CLI. The first AI-ERD tool call triggers the OAuth flow.

Gemini CLI's MCP support and config schema is evolving — consult the Gemini CLI docs for the current version.

---

Try it

List my AI-ERD ERD documents.

Show me the schema of the orders table in ERD document <uuid>.

In ERD document <uuid>, add an order_items table linking orders and products with quantity and unit_price columns. Use description "Add line items".

---

Tips

• Use the ERD document UUID in scripted prompts to avoid name resolution.
• Description matters — every write needs one and it shows up in your version history.
• Combine with Gemini's code generation — ask Gemini to read the schema first, then generate Go/Python/TypeScript model code that matches.

---

Troubleshooting

Gemini doesn't list AI-ERD tools Check ~/.gemini/settings.json is valid JSON and Gemini CLI was restarted after edits.

Auth flow fails Ensure your browser can open the AI-ERD authorization URL; if not, copy/paste the URL manually.

Tool call returns 403 Confirm your account has access to the target document (OWNER or EDITOR for writes).