by levi-qiao
Credential-scoped AI quota, context, and cache in Herdr for Claude, Codex, Grok, Agy, OpenCode, Pi, omp, and Devin.
# Add to your Claude Code skills
git clone https://github.com/levi-qiao/herdr-agent-quotaGuides for using ai agents skills like herdr-agent-quota.
Last scanned: 9/3/2026
{
"issues": [],
"status": "PASSED",
"scannedAt": "2026-09-03T08:35:01.283Z",
"npmAuditRan": true,
"pipAuditRan": true,
"promptInjectionRan": true
}herdr-agent-quota is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by levi-qiao. Credential-scoped AI quota, context, and cache in Herdr for Claude, Codex, Grok, Agy, OpenCode, Pi, omp, and Devin. It has 74 GitHub stars.
Yes. herdr-agent-quota passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.
Clone the repository with "git clone https://github.com/levi-qiao/herdr-agent-quota" and add it to your Claude Code skills directory (see the Installation section above).
herdr-agent-quota is primarily written in Rust. It is open-source under levi-qiao on GitHub, so you can review or fork the full source.
Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh herdr-agent-quota against similar tools.
No comments yet. Be the first to share your thoughts!
⚠️ Third-Party Software Notice
This skill is third-party open-source software developed and hosted independently on GitHub. SkillsLLM is an informational directory and does not control or maintain the underlying repository.
Any security checks, ratings, or warnings displayed by SkillsLLM are automated and limited in scope. They do not constitute a security certification or guarantee that the software is safe, error-free, or free from malicious code, vulnerabilities, compromised dependencies, or prompt-injection risks.
Review the source code, permissions, dependencies, and configuration before installing or running any third-party skill. Use is at your own risk. To the maximum extent permitted by applicable law, SkillsLLM is not liable for losses arising from third-party software.
Model, context, prompt-cache usage, and subscription quota in Herdr's Agent sidebar.
The plugin preserves Herdr's native machine/workspace/tab row, custom styles,
and worktree grouping. The branded provider/model line is the agent identity;
the native agent row is omitted so grok does not sit above Grok/grok-4.6.
Optional quota ordering and low-quota notifications are disabled by default.
Empty fields collapse; percentages can show remaining or used quota.
Requires Herdr 0.9.0+, the Rust toolchain pinned in rust-toolchain.toml,
macOS or Linux, and a supported agent CLI.
git clone https://github.com/levi-qiao/herdr-agent-quota.git
cd herdr-agent-quota
./install.sh
To enable a subset, use ./install.sh --agent claude,codex,omp.
Existing sessions need restarting only when newly installed hooks or Herdr
integrations must be loaded.
Upgrade from the repository directory:
git pull --ff-only
./install.sh
Upgrades retain saved preferences, repair managed configuration, refresh quota, and restore background updates automatically. No cache deletion or watcher management is required. Changes to the Herdr server connection are adopted by the watcher automatically.
Press prefix+shift+q, or run the following if that key is already assigned:
herdr plugin pane open --plugin herdr-agent-quota --entrypoint settings --focus
| Setting | Options |
|---|---|
| Percentages | Remaining or used; colors always indicate remaining headroom |
| Layout | packed groups related fields; stacked gives each field a row |
| Row gap | Zero or one blank line between agents |
| Watch interval | 30 seconds–1 hour; default 60 seconds |
| Fields | Topic, model, cache, TTL, context, short/long quota |
| Brand colors | On or off |
| Agent order | Herdr default or lowest remaining quota first |
| Low quota alert | Off or a threshold from 1% to 100% |
| Agents | Claude, Codex, Grok, Agy, OpenCode, Pi, OMP, Devin |
Use arrows or Space to edit, a to apply, and q to close.
Installer options are also available through ./install.sh --help.
| Agent | Quota source | Attribution |
|---|---|---|
| Codex | Codex app-server; 5h and/or 7d | Current login in the plugin's CODEX_HOME |
| Grok | CLI billing endpoint; 7d or 30d | Current CLI credentials |
| Devin | CLI usage endpoint; 1d and 7d | Current CLI credentials |
| Claude Code | StatusLine; 5h and 7d | Exact session observation |
| Agy / Antigravity | StatusLine; 5h and 7d | Exact session and identifiable model pool |
| OpenCode | OpenCode Go usage endpoint | Go credential; confirmed PAYG routes have no subscription quota |
| Pi | Canonical Codex quota | Only when the recorded account matches |
| OMP | omp usage --json --provider <id> |
Reported account matching the session's credential pin |
Quota windows retain their provider's meaning. Model, context, and cache data
come from the identified session when available. ttl≈ marks an estimated
prompt-cache lifetime, not a guaranteed expiry. Topic extraction uses only the
named pane's visible screen and preserves the last topic when it scrolls away.
All supported working agents participate in one background watcher. Requests are debounced for 60 seconds, including a final refresh after a turn settles. OMP additionally retains its own five-minute usage cache. Idle panes sharing a verified quota source receive the same reading.
Native Codex, Grok, and Devin collectors follow the plugin's current login, not separate accounts for each pane. Claude/Agy do not report a reliable serving account ID, so their observations are not shared across sessions. Unknown identity or model-pool attribution does not produce a guessed quota. Failed requests preserve the last verified reading for that same account; they do not turn failures into zero usage.
| Symptom | Check |
|---|---|
| Session data is missing | Run herdr integration status; load missing integrations before restarting the affected agent |
| Claude/Agy quota is missing | Send a turn so the session's StatusLine produces an observation |
| OMP quota is missing | Check omp usage --json --redact --provider <id> |
| Devin quota is missing | Check the CLI login and DEVIN_CREDENTIALS_FILE if customized |
| Rows are missing | Run the configure action below to repair managed configuration |
| Packed rows are truncated | Select stacked |
herdr plugin action invoke refresh --plugin herdr-agent-quota
herdr plugin action invoke configure --plugin herdr-agent-quota
Uninstall everything with ./uninstall.sh, or remove a subset with
./uninstall.sh --agent grok. Configuration changes are reversible; user-owned
settings and other agents remain intact.
See CONTRIBUTING.md for development and validation, SECURITY.md for data handling and vulnerability reports, and CHANGELOG.md for release notes. Dated investigations are indexed in docs/README.md.
MIT. Not affiliated with Herdr or the supported AI providers.