Configure AI Settings

What this setup controls

Curoxi uses your company OpenAI settings in the meeting automation flow. When matching meeting-note emails are pulled in through IMAP, Curoxi sends that content to the configured model and creates structured draft tasks for the selected project.

This is a company-level setting, so you should configure it once with the same care you use for billing or email credentials.

• Meeting note ingestion into customer workspaces
• Draft task creation from AI-generated notes
• AI-assisted workflow actions used in the workspace

Before you start

You need a company admin login in Curoxi, plus an OpenAI account that can create API keys and use API models. If you do not have access to these settings, contact support.

• Sign in as a company admin
• Open Settings and go to Meeting Automation or the AI configuration section
• Make sure your company IMAP setup is ready if you plan to use note-to-task automation
• Keep your OpenAI API key ready before you save changes

Step 1: Create an OpenAI API key

You create the key in the OpenAI platform, not inside Curoxi. Use a workspace or billing account that is approved for API usage.

1. Sign in to the OpenAI platform at platform.openai.com.
2. Open the API keys page.
3. Create a new secret key.
4. Copy the key immediately and store it safely. OpenAI will not show the full secret again after creation.
5. Use a company-owned key instead of a personal employee key whenever possible.

OpenAI API keys page

OpenAI models overview

Step 2: Save the key in Curoxi

• If a key already exists, Curoxi lets you leave the field blank to keep the current saved key.
• The model field is text-based, so the model name must be typed exactly.

1. In Curoxi, open Settings.
2. Go to the company Meeting Automation or AI settings area.
3. Paste the OpenAI API key into the OpenAI API Key field.
4. Enter the model name exactly as provided by OpenAI.
5. Save the settings.

Our recommended model

For most Curoxi customers, we recommend starting with gpt-5-mini. It is a practical default for note-to-task workflows because it balances cost, speed, and enough reasoning for structured task extraction.

If you want higher-quality reasoning for more complex meeting notes, you can move up to gpt-5.1. If your priority is the lowest possible cost for simple summarization, you can test gpt-5-nano.

• Recommended default: gpt-5-mini
• Higher-quality option: gpt-5.1
• Lowest-cost option: gpt-5-nano

Valid model names you can configure

Checked against OpenAI official model docs on April 27, 2026. These are current API model IDs that fit text-based Curoxi automation use cases.

• gpt-5.1
• gpt-5
• gpt-5-mini
• gpt-5-nano
• gpt-4.1
• gpt-4.1-mini
• gpt-4.1-nano

How to choose the right model

• Use gpt-5-mini if you want the best starting point for most customers.
• Use gpt-5.1 if your notes are long, messy, or need more careful task extraction.
• Use gpt-5-nano if cost matters more than depth and your inputs are short and consistent.
• Use the GPT-4.1 family only if you prefer a non-reasoning model path or already standardize on it elsewhere.

Step 3: Test the setup

• Start with a short test email before using live customer notes.
• If the output is too shallow, try gpt-5.1.
• If the output is good and cost matters, stay on gpt-5-mini or test gpt-5-nano.

1. Save the AI settings first.
2. Confirm the IMAP inbox configuration is active.
3. Select the project that should receive generated tasks.
4. Send one small test meeting note email through the configured inbox flow.
5. Check whether Curoxi creates draft tasks in the expected customer workspace and project.

Common setup mistakes

• Using a ChatGPT-only account without API access
• Pasting the wrong key or an expired key
• Entering a model nickname instead of the real API model ID
• Saving the OpenAI settings but not completing IMAP or project selection
• Expecting automation to work before sending a real test note

Explore Curoxi

Browse the core pages that explain how Curoxi works, how pricing is structured, and how to start a workspace.