Shotoku launches local-first budget caps and approval gates for AI agent spending

Shotoku is an open-source, local-first tool that enforces budgets, approval queues and a tamper-evident audit trail before an AI agent spends money. Policy lives in one YAML file with per-vendor maxAmount and maxDailyAmount limits, and every decision is written to a local data/decisions.jsonl file with no account, cloud or telemetry. It integrates via a TypeScript authorize() call or an MCP wrapping mode that places Shotoku in front of the agent's other MCP servers, including PayPal's, and the team says it is in beta and runs end to end against PayPal's sandbox.