autorun

agent-limit-retry Pro — subagent resume after a usage limit, for Claude Code, Codex and ZCode.

WHAT YOU GET • Full source code as a zip (with SHA-256), installed on top of the free MIT version: npm i -g the zip, done. No license server, no telemetry, nothing leaves your machine. • Subagent resume: after a 5-hour or weekly limit resets, the main agent is told exactly which subagents the limit killed (id, task, files changed, last steps) and to wake them by id instead of dispatching them again. Claude Code (SendMessage), Codex (followup_task), ZCode (SendMessage / ReadSessionContext). • Codex auto-continue: a per-thread watcher waits for the reset and queues "continue" with codex queue. Works with Codex CLI and Codex Desktop, macOS and Windows. • alr run --agent codex: unattended Codex runs that sleep through the reset and resume the same thread. • 12 months of updates (new zip by email), and the free version's features: pre-limit handoff, alr run for Claude Code, alr status, alr selftest, alr report.

REQUIREMENTS Node.js 18+. Claude Code 2.1.234+ on Pro/Max, Codex CLI/Desktop, or ZCode 3.14+. macOS verified end to end; Windows (Codex) reported by early testers.

DELIVERY By email within 24 hours of purchase, to the address you check out with. Reply to that email with your GitHub username if you want access to the private repo as well.

LICENSE & REFUND One license per person; use and modify it on every machine and account you own. Do not publish or resell the source. Full refund within 14 days, for any reason: email support@autorun.fun with your order number.

It cannot raise your limits or rotate accounts. It only helps you use the plan you already have.

Production-safe execution harness for AI agents. Full source code for Spring Boot 3.5 / Java 21 / PostgreSQL 16. No telemetry, no license server, runs entirely inside your infrastructure.

What you get

• Tool Gateway with strict schema validation and a sealed result model
• YAML Policy Engine: single-action limits, structured context rules, kill switch, fail-closed
• Intent / Attempt idempotency enforced by database constraints, three-state recovery (Applied / ConfirmedNotApplied / Indeterminate), bounded retries, crash-recovery scanner
• Approvals bound to immutable intents with evidence, content hash and an eight-step execution-start transaction; REST + minimal approval UI
• Append-only audit event stream (DB trigger), redaction, trace replay
• Shadow mode with human decision recording and per-tool agreement stats
• Purchase-agent example with a chaos supplier, 83 fault-injection tests on a real PostgreSQL 16, five runnable scenario scripts
• Six design notes, SECURITY.md, threat model

Delivery: full source delivered on September 21, 2026 as a zip with SHA-256 checksum, plus read access to the private update repository. 12 months of updates.

Refunds: full refund anytime before delivery. After delivery: refund if the product cannot run as described in the README and we cannot resolve it within 7 days.

License: use in any number of your own projects, including client work delivered as applications; do not publish, resell or redistribute the source. Full terms: https://autorun.fun/terms · License: https://autorun.fun/license · Privacy: https://autorun.fun/privacy · Support: support@autorun.fun