ChatGPT Onboarding

Why this matters: ChatGPT usage across your organization draws from a shared pool of credits or is billed in dollars, but the total doesn't show who is using it, on which models, or in which product. Connecting your ChatGPT workspace here brings that cost into Mavvrik's AI Assistant dashboards, alongside the rest of your AI spend.

Supported plans: Only workspaces on the ChatGPT Enterprise plan are supported today. The ChatGPT Business plan is not supported.

Before onboarding, make sure you are a workspace Owner and follow the steps below. Owners can create an admin key with custom permissions; admins only get the permission categories available to them.

Configuration steps

Step 1: Get your Organization ID and Workspace ID

Open the ChatGPT Admin Console and copy both IDs. Keep them handy for the Mavvrik onboarding form in Step 3.

Organization ID: starts with org-
Workspace ID: the ID (a UUID) of the workspace you want to connect

Step 2: Create a read-only admin API key

Sign in to the OpenAI Admin Console and select the workspace you want to connect
Go to Credentials → Admin keys → Create new admin key
Set the name to mavvrik-chatgpt-connector and choose Custom permissions
Set these permissions to Read and leave every other category at None:
- Enterprise analytics API: Read
- Users: Read
- Spend controls: Read
- Directory: Read
- Codex Enterprise analytics: Read
- Compliance logs: Costs and Workspace analytics only
Click Create and copy the key. It starts with sk-admin- and you will not be able to see it again

Permission	What Mavvrik uses it for
Enterprise analytics API: Read	Daily credits and cost per user, by model and product
Users: Read	Member names and emails, so cost shows people instead of IDs
Spend controls: Read	Each user's monthly spend limit, to show limit utilisation
Directory: Read	Groups and their members, to break cost down by group
Codex Enterprise analytics: Read	Codex credits and cost by type of work
Compliance logs: Costs	OpenAI's organization cost records, to reconcile cost
Compliance logs: Workspace analytics	Workspace usage activity, for activity dashboards

Step 3: Onboard in Mavvrik

Go to the Menu
Click on Admin → Accounts → GenAI Accounts
Click the + Account button
In the Select Provider dialog, under AI Assist, select ChatGPT
Choose Plan Enterprise, then enter the Organization ID and Workspace ID from Step 1, an optional Workspace Name, and the Admin API Key from Step 2
Click Connect

Mavvrik checks each permission separately and tells you exactly which one is missing, if any.

After onboarding is completed, allow up to 24 hours for data ingestion. Once ingestion is complete, cost will be visible under GenAI > Assistants > Cost.

Troubleshooting

"The admin API key is invalid or revoked": copy the key again, or create a new one, and make sure it starts with sk-admin-.
"Workspace ID not found for this admin key": check the Workspace ID, and make sure the key was created in that workspace's organization.
"The admin API key is missing the '…' permission": in the OpenAI Admin Console, edit the key under Credentials → Admin keys and set the named permission to Read.

FAQs

Which ChatGPT plans are supported?

Only workspaces on the ChatGPT Enterprise plan are supported today. ChatGPT Business appears in the Plan list but cannot be selected yet.

How is ChatGPT Enterprise billed?

ChatGPT Enterprise contracts are either credit-based or token-based. On credit-based contracts, advanced features such as Thinking models, Deep Research, image generation and Codex draw from a shared pool of credits, and usage past the pool bills as overage. On token-based contracts, usage is charged in US dollars at your contracted rates.

Which cost does Mavvrik show?

The cost OpenAI reports for each user and day. On credit-based contracts that is OpenAI's estimated dollar value of the credits used, at your workspace's credit rate; on token-based contracts it is the dollar amount OpenAI reports. Neither is the final invoice amount, and per-seat fees are not included.

What does Mavvrik's ChatGPT integration do?

It brings ChatGPT cost into Mavvrik and makes it analyzable by user, day, model, product (ChatGPT, Codex, Agents) and client (web, desktop, mobile, IDE), with each user's spend-limit utilisation. Engineering, FinOps and finance work from the same data instead of reconciling usage by hand.

Does Mavvrik read our conversations?

No. The admin key only has read access to usage and cost analytics, users and groups, spend limits, and the cost and workspace analytics records. It cannot read conversations, prompts, files or other compliance logs.

We have more than one workspace. How do we connect them?

Add each workspace as its own account in Mavvrik, each with its own admin key.

How do I rotate or revoke the key?

Create a new key with the same permissions in the OpenAI Admin Console, update the Admin API Key on the account in Mavvrik, then revoke the old key under Credentials → Admin keys.

Does Mavvrik replace OpenAI billing?

No. OpenAI remains the billing and product source. Mavvrik adds attribution, reporting, allocation and spend-limit visibility on top of your ChatGPT usage.