monday.com's AI features let you run prompts against monday-hosted AI models without managing your own model provider or API keys. The run_prompt mutation is a simplified, GraphQL-native entry point to monday.com's AI gateway: send a prompt, optionally tune the model and generation settings, and get the generated text back in the same request as the rest of your integration.

Requirements

Before you can use run_prompt, make sure the following are in place:

• External AI gateway access permission must be enabled for the account. Without it, the mutation returns a permission error.
• Apps: if you call run_prompt through an app (with an app token), the app must request the AI:Consume scope.
• AI token consumption: like the Models API, usage draws down monday AI tokens. The same metering and consumption rules apply. See Pricing and metering for details.

Mutations

Required scope (apps): AI:Consume

Run a prompt

Runs a single prompt against a monday-hosted AI model and returns the generated text. Returns a RunPromptResult.

mutation {
  run_prompt(
    prompt: "Summarize the benefits of regular exercise in two sentences."
    config: {
      model: MONDAY_STANDARD
      system_prompt: "You are a concise assistant."
      temperature: 0.7
      max_tokens: 200
    }
  ) {
    content
  }
}

The minimal form only requires a prompt:

mutation {
  run_prompt(prompt: "Say hello in exactly three words.") {
    content
  }
}

import { ApiClient } from "@mondaydotcomorg/api";
const mondayApiClient = new ApiClient({ token: myToken });

const query = `
  mutation {
    run_prompt(
      prompt: "Summarize the benefits of regular exercise in two sentences."
      config: { model: MONDAY_STANDARD, temperature: 0.7, max_tokens: 200 }
    ) {
      content
    }
  }
`;
const response = await mondayApiClient.request(query);

Arguments

Fields