Yes—ForgeBot can consume the same Codex & Work resources used in ChatGPT. The important distinction is which account and workspace fund the work, not whether the request starts in Slack or the ChatGPT app.
“You’re out of Codex and Work usage”
“Your workspace is out of credits. Add credits to continue using Codex and Work.”
The second line matters: this is a workspace-credit warning, not merely evidence that one person sent too many messages. The screenshot shows Work with GPT-5.6 Sol Medium; it does not identify the workspace, account, credit history, or who used the credits.
For standard ChatGPT Business seats, eligible work uses included plan limits first. Codex and ChatGPT Work share agentic usage allowances with supported features such as Excel, PowerPoint, Word, and Workspace Agents.
Being members of the same team does not mean every person’s included allowance is one undifferentiated team bucket.
When included limits are reached, purchased workspace credits can fund eligible additional use. This is a shared workspace pool. Codex-only usage-based seats require workspace credits from first use.
Work by one account can therefore reduce the credit balance available to other members, even when their individual allowances are separate.
Do not read this warning as “all ChatGPT is exhausted.” Ordinary Chat has separate model allowances and metering. Some paid Chat features can also consume workspace credits under flexible pricing, but regular Chat and Codex/Work are not one universal message counter.
Consequently: if ForgeBot is authenticated to the affected workspace, its eligible usage can contribute to that workspace’s shared credit consumption. If it uses a different workspace or a personal subscription, it does not drain this workspace merely because the requester is on the team.
The affected workspace has not been matched to ForgeBot’s billing identity in this review. Current routing also does not prove the routing used for every earlier integration task.
Yes, if those runs were funded by the affected workspace. Integration testing often involves repeated model turns: inspect a result, choose the next action, retry, test, and verify. Long context, larger models, parallel agents, and Fast mode can increase usage. One Slack request can therefore consume much more than one short chat exchange.
That makes the testing a plausible contributor—not a proven cause. The screenshot contains no per-user or per-task breakdown, so it would be unfair to assign the exhaustion to one person or project.
No credit purchases, billing changes, or model changes were made for this review. Waiting for a reset may restore included usage where applicable; it does not replenish purchased workspace credits.
The linked private Slack DM was not accessible to the bot. This analysis uses the question supplied in the current conversation and the accessible screenshot, not unseen DM history. Public documentation establishes product behavior; a workspace usage audit is still needed to establish who consumed what.