Custom / Local Model Setup
Use Custom / Local when you run an OpenAI-compatible server such as a local model host or an internal gateway.
Configure it
Section titled “Configure it”- Open Settings > Models.
- Set Coding Provider to
Custom / Local. - Enter the exact model ID exposed by the server.
- Enter the OpenAI-compatible base endpoint, typically ending in
/v1. - Enter a token if the server requires one.
- Run Test Connection.
- Save Settings and send a simple AGENT request.
There is one route. Do not look for a Logistics provider, second endpoint, Use one model toggle, or clarification-budget setting.
Common failures
Section titled “Common failures”Test Connection fails
Section titled “Test Connection fails”- Confirm the server is running and reachable from the Mac running FileMaker.
- Check the base URL, port,
/v1path, token, and model ID. - Confirm the server implements the OpenAI-compatible endpoint Genie calls.
Basic chat works but AGENT tools fail
Section titled “Basic chat works but AGENT tools fail”The selected model or server may not support tool calling, required tool choice, or compatible structured responses. Choose a tool-capable model or gateway.
Genie works but native coding actions are unavailable
Section titled “Genie works but native coding actions are unavailable”Provider access and Genie product access are separate. The free version of Genie can answer and provide human-readable implementation guidance; native script artifacts, LAYOUT, DATA Execute, and external Local MCP require Coding Access.