SIwithin - Run Claude Code unattended with fences and budgets
Show HN: Run Claude Code unattended inside a fence, under a dollar budget
SIwithin lets you run Claude Code and OpenAI Codex command-line agents unattended on a brief, with three hard guarantees: the run only touches folders you explicitly allow, it cannot exceed a dollar budget, and it leaves a single ledger line per run. Using macOS sandbox-exec and a loopback proxy for the Claude adapter, it refuses your home folder, meters every API request, and kills the run if the budget is hit. Installation is a single Python file with no dependencies, making it easy to try: clone, set a cap, and run with --dry-run to preview everything.
It keeps three promises with a tool and not with hope: the run touches only the folders it was given, cannot spend more than the dollars it was given, and leaves one line saying what it cost and whether it was done.