hermes - 💡(How to fix) Fix Cron jobs and agent dispatches burn provider credits without user consent or budget cap [1 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
NousResearch/hermes-agent#23419Fetched 2026-05-11 03:29:34
View on GitHub
Comments
0
Participants
1
Timeline
3
Reactions
0
Author
Participants
Timeline (top)
labeled ×3

Fix Action

Fix / Workaround

Hermes cron jobs dispatch agents to LLM providers (xAI Grok, OpenRouter, etc.) automatically without any budget cap, spending confirmation, or per-job cost estimate. The user has no visibility into what a cron job will cost before it runs, and no mechanism to cap total daily/weekly spend.

In our case, Hermes was configured by a previous Claude session with:

  • 2 active cron scripts (critical_watch every 30 min, approval_queue refresh every hour)
  • 12 commented-out cron stubs that were never wired up
  • An agents-teach-agents script with 38 random agent pairs
  • A launch_autopilot script that auto-dispatched 3 agents per night
  • Default model set to grok-4-fast-reasoning via xAI
  1. No pre-dispatch cost estimate. Before a cron job fires, Hermes should estimate the token cost and either log it or gate it behind a configurable budget.
RAW_BUFFERClick to expand / collapse

Problem

Hermes cron jobs dispatch agents to LLM providers (xAI Grok, OpenRouter, etc.) automatically without any budget cap, spending confirmation, or per-job cost estimate. The user has no visibility into what a cron job will cost before it runs, and no mechanism to cap total daily/weekly spend.

In our case, Hermes was configured by a previous Claude session with:

  • 2 active cron scripts (critical_watch every 30 min, approval_queue refresh every hour)
  • 12 commented-out cron stubs that were never wired up
  • An agents-teach-agents script with 38 random agent pairs
  • A launch_autopilot script that auto-dispatched 3 agents per night
  • Default model set to grok-4-fast-reasoning via xAI

The result: ~$20 USD burned in a single day on xAI credits, mostly on agent reasoning tokens (not the image/video tools the provider was designated for). The user (a bootstrapped startup founder) discovered this only by checking the xAI billing dashboard manually. Hermes provided no warning, no confirmation, no cost estimate, and no budget cap.

Specific issues

  1. No pre-dispatch cost estimate. Before a cron job fires, Hermes should estimate the token cost and either log it or gate it behind a configurable budget.

  2. No daily/weekly/monthly spend cap. There is no max_daily_spend_usd or equivalent in config.yaml. A runaway cron can drain an entire provider balance.

  3. No per-provider budget isolation. The user designated xAI for image/video tools only, but Hermes routed agent reasoning through xAI's grok-4-fast-reasoning model. There is no way to say "provider X is for tools only, never for reasoning."

  4. Silent cron execution. Cron jobs run and succeed without any user-facing notification of cost. The PAUSE file mechanism exists but is manual. There is no automatic circuit breaker.

  5. Related to #4379 (73% token overhead per call) and #3015 (session file leak causing cost explosion): the pattern is that Hermes silently accumulates costs in ways the user cannot see or control until after the money is gone.

Expected behavior

  • config.yaml should support budget.daily_limit_usd, budget.monthly_limit_usd, and per-provider limits
  • Cron jobs should log estimated cost before execution
  • When a budget limit is hit, all dispatches should pause and notify the user
  • Provider designation should support "tools-only" mode (use this provider for specific tool calls, never for agent reasoning)

Environment

  • Hermes Agent v0.11.0 (2026.4.23)
  • Provider: xAI (grok-4-fast-reasoning)
  • OS: Ubuntu Linux
  • Config: default config.yaml with xAI as primary provider

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…

FAQ

Expected behavior

  • config.yaml should support budget.daily_limit_usd, budget.monthly_limit_usd, and per-provider limits
  • Cron jobs should log estimated cost before execution
  • When a budget limit is hit, all dispatches should pause and notify the user
  • Provider designation should support "tools-only" mode (use this provider for specific tool calls, never for agent reasoning)

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING