Skip to content

Recipes

Structured, repeatable playbooks for humans and agents -- each recipe states when to use it, what it needs, and the exact steps to run against LLM Prover.

MCP Recipes

Compare Models Over MCP

Fire one prompt at several models through the LLM Prover MCP server, wait for the run to finish, and read the structured comparison.

Use when: An agent needs side-by-side output from multiple models for the same prompt, driven entirely over MCP with no human in the loop.

Connect an MCP Client

Point an MCP-compatible agent or IDE at the LLM Prover server and confirm the tool list loads.

Use when: Setting up a new agent or IDE against the MCP server for the first time, before running any tool.

Track a Run Across Sessions

Kick off a long-running LLM Prover job, remember it with an agent-state note, and reconcile it in a later session instead of losing track.

Use when: A job takes a while (or you want to close your laptop and come back), and you don't want to lose track of which run you were waiting on.

Monitor a Model for Drift Over Time

Stand up a scheduled benchmark, then have your agent reconcile each run against the baseline and tell you only when quality, cost, or latency moves.

Use when: You have picked a model (or pipeline) for a task and want to be told if it regresses over time -- a quality drop, a cost spike, a latency creep -- without watching a dashboard.

Reconcile Your Runs on Startup

Set up a standing instruction so a new agent session automatically checks what it was watching and reports any LLM Prover runs that finished while you were away.

Use when: You run multi-session recipes (drift monitors, long jobs) and want each new session to pick up where the last left off, without you having to ask.