OpenAI Adds API-Key Spend Dashboards and Hard Limits
OpenAI's API platform now lets teams group Usage and Costs data by API key and set monthly organization or project spend limits. The dashboard feature helps identify which apps and workloads drive spend, while a hard limit can stop affected API traffic with a documented 429 error after the cap is reached.
OpenAI says teams can now track API usage and spend by API key in the Usage and Spend dashboards, then set monthly organization or project spend limits, including hard limits that stop affected traffic when reached. (OpenAI Developers on X)
The update addresses two different operations questions: which app or workload used the budget, and what should happen when the budget is exhausted. OpenAI's durable documentation confirms API-key grouping in the Usage and Costs dashboards and APIs, plus hard-limit behavior for organizations and projects. (OpenAI API changelog)
See which API keys drive usage
OpenAI's August 4 changelog entry says customers can filter and group data by API key in the Usage and Costs dashboards. The Usage API and Costs API also support the API-key dimension for programmatic reporting and analysis. (OpenAI API changelog)
That gives teams a more useful attribution layer than a single organization-wide total. A service can keep a separate key for a production backend, an internal prototype, or a scheduled batch job, then compare those keys in the dashboard or pull the same dimension into its own reporting pipeline. The cited documentation confirms the grouping capability, but does not promise that every existing key is already named or separated in a particular workspace.
A practical first pass is to map each key to one owner, app, or environment before reviewing the data. If several unrelated services share one key, the dashboard can show that key's activity but cannot reconstruct the attribution that was never separated at issuance.
Set a monthly organization or project cap
OpenAI's spend-limit guide separates notification from enforcement:
| Control | What happens at the configured amount | Use it when |
|---|---|---|
| Spend alert | Sends a notification; API traffic continues | You want time to investigate or adjust usage |
| Hard spend limit | Affected API requests return a 429 error | You need a monthly organization or project cap |
An organization hard limit applies to API traffic across all projects in the organization. A project hard limit applies only to API traffic billed to that project. Both can apply to a request, so a project that has room remaining can still be affected if the organization-level limit has been reached. (OpenAI spend limits)
The changelog dates the organization and project hard-limit feature to July 22, 2026. It says a monthly cap causes affected API requests to return a 429 error when tracked spend reaches the limit, and recommends spend alerts for notification before traffic is interrupted. (OpenAI API changelog)
Know the error codes and reset path
OpenAI documents separate error codes for the two scopes:
organization_spend_limit_exceededmeans the organization hard limit was reached.project_spend_limit_exceededmeans the project hard limit was reached.
The spend-limit guide says raising or removing the reached limit allows traffic to resume after the update propagates. Otherwise, the limit resets with the next monthly cycle. When an affected request fails, inspect error.code instead of treating every billing-related 429 as a token rate-limit event. (OpenAI spend limits)
There is an important operational caveat: enforcement is not instantaneous. OpenAI says the API Platform can process a small amount of extra usage while the limit state propagates, which means recorded spend can slightly exceed the configured amount. A hard limit is therefore a safety control, not a perfect transaction boundary. (OpenAI spend limits)
A safer rollout pattern
Start with API-key attribution before turning on enforcement. Group Usage and Costs data by key, identify the services that should share an organization or project budget, and add spend alerts below the cap so the team has time to respond. The dashboard and API-key dimension are documented for analysis; the hard-limit controls are documented separately for interruption. (OpenAI API changelog)
Then set a project limit for a contained workload before applying an organization-wide hard limit. The project scope narrows the blast radius, while the organization scope is useful when every project must share a firm monthly ceiling. In both cases, document who can raise the limit and how the service should handle the corresponding error code.
Do not confuse the OpenAI-approved usage limit for an organization with a configured spend limit. OpenAI says the approved monthly usage limit is separate from spend limits that an organization or project administrator configures. (OpenAI spend limits)
What this update does not change
The cited documentation does not describe API-key grouping as a permission system or a way to retroactively split shared-key activity between applications. It also does not make spend alerts enforce a cap: alerts notify, while hard limits can return errors. (OpenAI API changelog; OpenAI spend limits)
The useful mental model is simple: API-key dimensions explain the bill, and spend limits constrain the bill. Use the first to find the workload that needs attention, then use alerts or a hard limit according to how much interruption your system can tolerate.
Sources
- OpenAI Developers on X, "You can now track usage and spend by API key...", August 21, 2026: https://x.com/OpenAIDevs/status/2090903221636338057
- OpenAI API changelog, August 4 and July 22, 2026 entries: https://developers.openai.com/api/docs/changelog
- OpenAI, Spend limits: https://developers.openai.com/api/docs/guides/spend-limits
- OpenAI, Admin APIs: https://developers.openai.com/api/docs/guides/admin-apis
Read next
Keep building the workspace playbookCursor adds enterprise model controls, soft spend limits, and richer usage analytics
Cursor's May 4, 2026 update adds granular model/provider access controls for Enterprise admins, introduces soft spend limits with automated alerts, and expands usage analytics so admins can break consumption down by product surface (including Cloud Agents, Bugbot, and Security Review).
OpenAI's Secure MCP Tunnel: Connect Private MCP Servers Over Outbound-Only HTTPS
OpenAI's Secure MCP Tunnel connects private and on-prem MCP servers to ChatGPT, Codex, and the Responses API without opening inbound firewall ports. You need a tunnel_id, an API key with Tunnels Read plus Use, and -- separately -- ChatGPT developer-mode access. That permission split is the most common setup blocker.
OpenAI Releases Codex Security CLI for Repository Scans and CI Checks
OpenAI says its open-source Codex Security CLI can scan repositories, track findings across runs, verify fixes, and add security checks to CI/CD. The beta CLI requires Codex Security access and is built for teams that want code-aware security review in the terminal.
Frequently Asked Questions
What can OpenAI API users see by API key?
OpenAI says the Usage and Costs dashboards can filter and group data by API key. The Usage API and Costs API also support an API-key dimension for programmatic reporting and analysis.
What happens when an OpenAI hard spend limit is reached?
Affected API requests return a 429 error. OpenAI documents organization_spend_limit_exceeded and project_spend_limit_exceeded error codes for the corresponding limits.
Is a spend alert the same as a hard spend limit?
No. A spend alert sends a notification while API traffic continues. A hard spend limit enforces a monthly cap and can interrupt affected API traffic after the limit is reached.
Can OpenAI spend limits stop production traffic immediately?
They can interrupt production traffic, but OpenAI says enforcement is not instantaneous. A small amount of extra usage can be processed while the limit state propagates, so recorded spend may slightly exceed the configured amount.