Cursor IDE includes native MCP support, meaning apex-mcp’s 116 Oracle APEX tools are available directly inside the Chat panel alongside any model you have configured — Claude, GPT-4o, or others. Once connected, you can create applications, generate CRUD pages, add JET charts, and inspect your schema without leaving the editor.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/TechFernandesLTDA/apex-mcp/llms.txt
Use this file to discover all available pages before exploring further.
Install apex-mcp
Clone the repository and install it as an editable package:Confirm the installation:
Create .cursor/mcp.json
Create (or edit) the file Replace every Alternatively, you can add the server through the Cursor UI: go to Cursor Settings → MCP → Add new global MCP server and paste the
.cursor/mcp.json at the root of your project. Cursor automatically reads this file for project-scoped MCP servers.YOUR_* placeholder with your actual Oracle and APEX credentials. Set cwd to the absolute path of the apex-mcp directory, or use ${workspaceFolder} if you opened apex-mcp itself as the workspace.Project vs. global config. The
.cursor/mcp.json file applies only to the current project. For a global config that applies across all projects, use ~/.cursor/mcp.json with the same format and an absolute cwd path.apex-mcp block.Restart Cursor
Fully close and reopen Cursor so it picks up the new MCP configuration. Cursor starts the
apex-mcp process in the background when the project loads.Verify tools are available in the AI panel
Open the Chat panel with ⌘L (macOS) or Ctrl+L (Windows/Linux) and switch to Agent mode. You can verify the MCP server is connected by asking:Cursor will display the tools it found from apex-mcp. You can also inspect MCP server status and logs via View → Output → MCP.
If the server shows
error status in the MCP output, verify that python resolves correctly in your shell. If you use a virtual environment, replace "command": "python" with the full path to the venv interpreter, e.g. "/home/user/apex-mcp/.venv/bin/python".Choosing the Right Model
apex-mcp works with any model available in Cursor — the tools are model-agnostic. However, complex multi-step APEX builds involve many sequential tool calls, so models with larger context windows handle them better.| Model | Suited for |
|---|---|
| Claude 3.5 Sonnet / 3.7 Sonnet | Full app generation, multi-step builds, complex schema inspection |
| GPT-4o | General APEX tasks, CRUD generation, page updates |
| Smaller / faster models | Single-tool queries, quick lookups, status checks |
Troubleshooting
| Problem | Solution |
|---|---|
Server shows error in MCP output | Check View → Output → MCP for the exact error message |
python: command not found | Use the full interpreter path: "command": "/usr/bin/python3" or your venv path |
| Tools appear but fail with Oracle errors | Verify all env values — wallet path must be absolute and contain cwallet.sso |
| Works with Claude model, not GPT | Some models handle large tool schemas differently; try rephrasing the request more directly |
| Config not picked up after editing | Restart Cursor fully; changes to .cursor/mcp.json are not hot-reloaded |
