Which Plan Powers ForgeBot?
Hermes is using ChatGPT Pro through Codex OAuth. The account has 64% of its weekly allowance remaining. A standard Business seat does not include the same allowance as Pro $200.
2:21 PM PDT · 21:21 UTC
The Important Catch
The primary model uses the subscription, but Hermes still has a paid OpenRouter fallback configured. If triggered and usable, that fallback can incur metered API charges. The current setup is therefore not a verified subscription-only configuration.
No billing settings, credentials, fallback settings, or banked resets were changed during this audit.
What Was Verified Live
- Default and current model: GPT-6 Astra, through Hermes’s
openai-codexprovider and ChatGPT’s Codex backend. - Identity: the bot email Adam named matches the OAuth credential’s verified-email claim. The full email and all credential material are omitted from this link-accessible report.
- Plan: both the credential claim and live usage service report
pro. They do not distinguish Pro $100 from Pro $200. Adam’s expected $200 plan is plausible but the invoice or My Plan page is still needed to confirm the exact tier. - Allowance: the service reports one seven-day window, 36% used, and no additional five-hour window in this response. That is not proof that every feature is unlimited.
- Extra usage: no purchased credits; balance 0. One reset is banked, but the service reports zero currently applicable resets. It was not redeemed.
Snapshot: September 23, 2026, 12:13 PM HST / 3:13 PM PDT. Usage can move while this report is being read. The displayed reset is the account’s reset time, not a presumed Sunday or Monday boundary.
Pro Versus A ForgeFX Team Member
“Team” is the older product name for ChatGPT Business. The comparison below assumes ForgeFX’s workspace is Business, rather than Enterprise. Its actual seat assignments and contract were not inspected.
| Plan / seat | Included Work & Codex allowance | What that means |
|---|---|---|
| Pro $200 Individual subscription | 20× Plus usage | Highest of these published included-usage tiers. Existing account billing must confirm this is the bot’s tier. |
| Pro $100 Individual subscription | 5× Plus usage | Same broad Pro capabilities, smaller allowance than Pro $200. |
| Business Standard Per-member seat | Plus-level published Codex estimates | $20/member/month annually or $25 monthly; minimum two users. Limited Astra usage within its allowance. |
| Business Premium Per-member seat | 5× Standard usage | Published $100 Business tier uses Pro 5× estimates. No five-hour limit; full existing allowance usable with Astra. This is not Pro $200’s 20× tier. |
Practical comparison: published usage tiers put Pro $200 at 20× a Standard seat and 4× a Premium seat. These compare included allowances—not guaranteed messages, speed, quality, or exact token quotas.
Each employee signs in with their own identity and uses the allowance attached to their selected workspace and seat. They do not borrow the bot’s Pro allowance. Purchased workspace credits are a separate shared pool that can fund additional usage when admin spending controls permit it.
Using Hermes instead of a Codex app does not create another subscription allowance. Activity under the same bot account draws from that account’s allowance across applicable clients and features.
There Is More Than One Usage Meter
Work And Codex
ChatGPT Work and Codex share usage. Applicable agentic features, including Excel and Workspace Agents where supported, can draw from shared allowances or credits too.
Model choice, task length, context, reasoning effort, speed and tools affect consumption. A long Hermes conversation can consume much more than a short prompt.
OpenAI’s illustrative local-message estimates for Astra are 5–45 for Standard Business, 25–225 for Pro 5×, and 100–900 for Pro 20× per five-hour comparison period. These are estimates, not fixed entitlements; live limits govern.
Ordinary ChatGPT Chat
Instant chat can be virtually unlimited subject to safeguards. That does not make reasoning, Pro models, Work, Codex or tools unlimited.
Business Standard includes 15 Pro-model Chat messages per month; Premium includes 50 per week, shared across GPT-6 Pro and GPT-5.6 Sol Pro. These Chat allowances are separate from Work and Codex.
The bot’s 64% reading is therefore not “64% of every ChatGPT feature,” nor the remaining allowance of any employee.
Keep Spending Flat-Rate
- Keep the primary route on Codex OAuth. It uses the ChatGPT subscription rather than an OpenAI Platform API key for this model.
- Remove or disable paid fallback routes in a separate configuration change. The detected fallback is OpenRouter with Claude Sonnet 4. A successful OAuth primary check does not close that spending route.
- Audit auxiliary models and external tools. Search, vision, image generation, speech and other integrations can have separate costs. No complete tool-billing audit was performed here.
- Avoid purchased overage credits and automatic top-ups. Subscription-backed extra credits are still additional spending, even though they are not raw API billing. The bot’s checked Codex balance is currently zero; workspace top-up settings were not verified.
- When included usage runs out, wait or use an eligible banked reset. Do not silently switch to API-key billing. Choose lighter subscription models for routine work when appropriate.
About the “10–100×” claim: heavy agent usage can make a flat subscription much cheaper, but that multiplier is not universal or verified here. A defensible comparison needs actual input/output/cached tokens, model rates and workload volume. API billing and ChatGPT subscriptions are separate; the subscription does not provide a general-purpose API budget.
Account governance: Business provides workspace controls and no model training on business data by default. Personal Pro uses the account’s data controls. Do not distribute the bot’s login as a substitute for employee seats; technical OAuth access alone does not establish approval for every shared-service use.
Open Questions And Next Decision
- Confirm the bot’s exact price tier in Settings → My Plan or its invoice. The checked usage endpoint only says Pro.
- Confirm whether each ForgeFX member has a Standard or Premium Business seat, and whether workspace credit spending is enabled.
- Prioritize disabling the paid fallback and reviewing auxiliary routing if “never pay raw API costs” is the hard requirement.
OpenAI currently says new sign-ups and upgrades to Pro $200 have been paused since September 10, 2026. Do not assume a canceled or downgraded $200 seat can immediately be restored.
Sources And Recheck
Live local evidence: configured default provider/model and fallback; OAuth identity claims (values not published); successful account-usage request. The usage service returned HTTP 200, plan Pro, 36% used, a seven-day window and reset time 2026-09-25T21:21:14Z.
Recheck on this host: hermes usage --provider openai-codex --json. A team member can check their own Codex usage dashboard or use /status inside Codex CLI.
- OpenAI: Pro tiers and $200 signup pause
- OpenAI: Business seats, models and limits
- OpenAI: Codex pricing and usage estimates
- OpenAI: Codex subscription access and shared usage
- Hermes: command and configuration reference
This is an unlisted, link-accessible working report, not a private portal. It contains no authentication tokens or account IDs. Plan documentation was checked September 23, 2026 and may change.