Schema | Houtini LM

Houtini LM

Describes the environment variables required to run the server.

Name	Required	Description
`HOUTINI_LM_MODEL`	No	Pin a specific model to use, since endpoints like OpenAI list dozens of models that all score the same in routing.
`HOUTINI_LM_API_KEY`	No	API key for the endpoint, if it needs one (e.g. your OpenAI key).
`HOUTINI_LM_THINKING`	No	Controls model thinking: 'auto' (default) switches thinking off for models detected as supporting the toggle, 'off' forces it off on every call, 'on' forces thinking on (useful for bug-hunting, at the cost of time and tokens).
`HOUTINI_LM_SERIALISE`	No	Set to 0 to turn off call queueing. Calls are queued one at a time by default; OpenRouter skips the queue, and for OpenAI, a router or a batching backend like vLLM you can turn it off.
`HOUTINI_LM_MIN_TOKENS`	No	The floor for the output token budget. Anything under 4,096 is ignored unless you set HOUTINI_LM_MIN_TOKENS=0, which allows tiny caps passed by MCP clients.
`HOUTINI_LM_STRUCTURED`	No	Set to 1 to have each inference result also carry a structuredContent block (answer, model, tokens, timing, quality flags) as JSON. Intended for driving houtini-lm from your own scripts rather than Claude Code.
`HOUTINI_LM_ENDPOINT_URL`	No	The URL of the OpenAI-compatible endpoint to delegate work to (e.g. http://192.168.1.50:1234, https://api.openai.com). Defaults to LM Studio on http://localhost:1234.

Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.

This server publishes no instructions, or was last inspected before Glama recorded them.

Server capabilities have not been inspected yet.

Functions exposed to the LLM to take actions

Name	Description
No tools

Interactive templates invoked by user choice

Name	Description
No prompts

Contextual data attached and managed by the client

Name	Description
No resources

This server cannot be deployed

ActivityActive

ResponsivenessNo issues