Shotoku
Spending limits, human approvals and an audit trail for AI agents, checked on your machine before a payment or paid API call goes through.
- AI tool
- Freemium
Pricing
Core is free and open source (MIT) and runs on your machine; Team and Enterprise plans for shared control are announced.
Overview
Shotoku gives AI agents a budget instead of your card. You write rules in one YAML file, such as a per-purchase and daily limit for each vendor, and the agent calls Shotoku before it spends money, calls a paid API, uses an MCP tool or runs a script. Each request is approved, denied, or held until a person approves it, and every decision goes into a tamper-evident log on your machine. It ships as a CLI, a TypeScript library and an MCP server for Claude, Cursor and other agents, and it never holds your funds. You get an allow, deny or wait-for-approval decision on every agent spend, with a record of each one.
Key features
- Per-vendor and daily spending limits
- Human approval for large spends
- Checks payments, paid API calls, MCP tools and scripts
- Tamper-evident decision log on your machine
- Policies in one YAML file
- CLI, TypeScript library and MCP server
- Open source under MIT
Use cases
- LLM spend limits
- Enforce spending limits
- Agent spending controls
- Approval workflows
- Agent permission approvals
Show all 9Show fewer
Alternatives
Last updated 2026-11-15