OpenAI

Connect your OpenAI account to Rootly to power AI-assisted incident workflows using your organization’s API key, data agreements, and preferred regional endpoint.

Introduction

The OpenAI integration lets you connect Rootly to your organization’s OpenAI account, using your own API key and Organization ID to power AI-assisted incident workflows. This gives your team control over model selection, data handling, and API usage under your organization’s specific agreements with OpenAI — including data retention policies that may differ from Rootly’s default OpenAI configuration.

With the OpenAI integration, you can:

• Generate AI-powered summaries, analyses, and responses within incident workflows
• Send custom prompts to GPT and reasoning models with full Liquid template support
• Use a system prompt to define the model’s role, tone, or output constraints
• Route requests through a regional OpenAI endpoint to meet data residency requirements
• Access OpenAI’s reasoning models (o1, o3) with configurable reasoning effort and summary output

Before You Begin

Before setting up the OpenAI integration, make sure you have:

• A Rootly account with permission to manage integrations
• An OpenAI API key with the following permissions:
  • Models permission set to Read
  • Model Capabilities permission set to Write
• Your OpenAI Organization ID

We recommend creating a dedicated API key for Rootly rather than using a personal key. Your key is validated on save and encrypted at rest in Rootly.

Installation

1. Open the integrations page in Rootly

   Navigate to the integrations page in your Rootly workspace and select OpenAI.

2. Enter your API key and Organization ID

   Paste your OpenAI API key and Organization ID into the respective fields. Rootly validates the credentials against the OpenAI API before saving.

   
3. Select a regional host

   Choose the OpenAI API host for your requests. The default is api.openai.com. If your organization requires data to stay within a specific region, select the appropriate regional endpoint:

   Host	Region
   api.openai.com	Default (US)
   us.api.openai.com	United States
   eu.api.openai.com	Europe
   gb.api.openai.com	United Kingdom
   ae.api.openai.com	UAE
   au.api.openai.com	Australia
   ca.api.openai.com	Canada
   jp.api.openai.com	Japan
   in.api.openai.com	India
   sg.api.openai.com	Singapore
   kr.api.openai.com	South Korea

4. Integration connected

   Your OpenAI integration is active. The Create OpenAI Chat Completion workflow action is now available in your incident and action item workflows.

Workflow Actions

Create OpenAI Chat Completion

Sends a prompt to an OpenAI model and captures the response as a workflow output. Supports both standard GPT models via the Chat Completions API and reasoning models (o1, o3) via the Responses API.

Field	Description	Required
Model	The OpenAI model to use — fetched from your account	Yes
Prompt	The user message — supports Liquid templating	Yes
System Prompt	Instructions for the model’s role or behavior — supports Liquid templating	No
Temperature	Sampling temperature between 0.0 and 2.0 — controls randomness	No
Max Tokens	Maximum number of tokens in the response	No
Top P	Nucleus sampling probability between 0.0 and 1.0	No
Reasoning Effort	For reasoning models: minimal, low, medium, or high	No
Reasoning Summary	For reasoning models: auto, concise, or detailed	No

Use Liquid variables in your prompts to include live incident context — for example {{ incident.title }}, {{ incident.severity }}, and {{ incident.description }}. See the Liquid variables reference for all available fields.

The System Prompt field sets the model’s persona or output format — for example: “You are an incident response assistant. Respond in bullet points. Be concise.”

Reasoning models (o1-*, o3-*) use the OpenAI Responses API instead of the Chat Completions API. When using these models, use Reasoning Effort to control how much compute the model uses before responding, and Reasoning Summary to control how much of that reasoning is surfaced in the output.

Troubleshooting

The API key or Organization ID is rejected on save

Rootly validates your credentials by making a test request to OpenAI. Confirm that the API key is active, has not been revoked, and has the correct permissions: Models set to Read and Model Capabilities set to Write. Also verify that the Organization ID matches the organization the API key belongs to.

The workflow action fails with an authentication error

If the integration was working and then stopped, the API key may have been rotated or revoked. Update the key in the integration settings — Rootly re-validates on save. Also confirm the Organization ID has not changed.

The workflow action fails with a rate limit error

OpenAI enforces rate limits based on your usage tier. Running many concurrent workflows may exceed requests-per-minute or tokens-per-minute limits. Consider staggering workflows, reducing token usage with more focused prompts, or upgrading your OpenAI tier.

A model is not appearing in the model selector

The model list is fetched dynamically from your OpenAI account. Rootly filters for gpt-*, o1-*, and o3-* models. If a model you expect to see is missing, confirm your API key has access to it — some models require specific OpenAI tiers.

Reasoning model parameters are being ignored

Reasoning effort and reasoning summary only apply to o1-* and o3-* models using the Responses API. If you select a standard GPT model, these fields are ignored and only the Chat Completions parameters (temperature, max tokens, top p) apply.

Liquid variables are not rendering correctly in the prompt

Check your Liquid syntax — unclosed tags or undefined variables can cause rendering failures. Use the Liquid variables reference to confirm variable names and test your template in a low-stakes workflow first.

Related Pages

Incident Workflows

Build workflows that use OpenAI models to analyze, summarize, or respond to incidents.

Liquid Variables

Reference for all incident variables available in Liquid-templated prompts.

Rootly AI

Learn about Rootly’s built-in AI features for incident management.