Shotoku

Open-source controls that let developers budget, approve, and audit AI agent spending.

✓ Verified against shotoku.dev · Nov 14, 2026Free tierCommand line · API

Visit shotoku.dev(opens in a new tab)
The Shotoku home page
Captured from shotoku.dev on Nov 14, 2026

What it does

Shotoku is an AI tool for governing AI agent actions and adding human approval to AI workflows, made by Julius (@juliuspsc) and Issa (@prunier_issa). It has a free plan.

Shotoku is an open-source tool for putting budgets, policies, and human approvals in front of AI agent actions, with a focus on spending. It is built for developers and teams whose agents call paid APIs, buy products, issue refunds, or use tools through MCP. Rather than giving an agent a payment method with no limits, the user writes rules that decide what is allowed. It works through an authorize() function in TypeScript or through an MCP server for compatible agents. Policies live in a YAML file and can set allowed actions per vendor, per-transaction limits, daily limits, and an overall budget for an agent. A request that goes over a limit is denied before the payment happens; a request that matches no rule waits for a person. Decisions are written to a local append-only JSONL ledger. Approvals can be handled from the terminal, a local dashboard, or Telegram. Shotoku holds no funds and stores no credentials, so the agent's own payment rail does the executing. In MCP wrapping mode, it can sit in front of other MCP servers so their tools are reached only through the policy check. The core is free and open source for use on the user's own machine. A Team plan is coming soon with organization-level policy hierarchy, Slack approval routing, multi-approver rules, and a central ledger dashboard. Enterprise adds signed decision receipts, role-based access control on approvals, compliance export packs, SSO/SAML/SCIM, and a support SLA.

  • Local policy engine for agent spending
  • Per-vendor transaction and daily limits
  • Human approval queue
  • Telegram, local dashboard, and terminal approvals
  • Append-only JSONL decision ledger
  • TypeScript authorize() API and MCP server
  • MCP wrapping for other tool servers
  • x402 payment authorization

Verified facts

Each row links to the page we checked it on.
APIYes, an authorize() function plus an MCP servershotoku.dev✓ Nov 14, 2026
Free planYes, a free forever plan on your own machineshotoku.dev✓ Nov 14, 2026
HeadquartersFranceshotoku.dev✓ Nov 14, 2026
MakerJulius (@juliuspsc) and Issa (@prunier_issa)shotoku.dev✓ Nov 14, 2026
Open-source licenseOpen sourceshotoku.dev✓ Nov 14, 2026
PlatformsCLI and APIshotoku.dev✓ Nov 14, 2026
PricingFree open-source core, with Team (coming soon) and Enterprise plansshotoku.dev✓ Nov 14, 2026
Shotoku pricing in detail →

What you can do with it

Good at

  • Runs on the user's machine without an account, cloud service, or telemetry
  • Fails closed when policy is missing or a request is malformed
  • Works with any agent that can call a function and with MCP-compatible agents
  • Open-source core with inspectable local decisions

Watch out for

  • Beta software, with an API expected to evolve
  • Team plan is not yet available
  • Slack approval routing is not yet offered

Common questions

What is Shotoku used for?
Shotoku is used for governing AI agent actions and adding human approval to AI workflows. On AIToolMaps it ranks #2 of 10 for governing AI agent actions.
Is Shotoku free?
Yes. Shotoku has a free plan.
What platforms does Shotoku run on?
Shotoku runs on Command line and API.
Does Shotoku have an API?
Yes. Shotoku offers an API, so developers can use it from their own code.
Who makes Shotoku?
Shotoku is made by Julius (@juliuspsc) and Issa (@prunier_issa).
What are the best alternatives to Shotoku?
The closest alternatives are Tess, Activepieces and Rigorix, which do most of the same jobs. Activepieces is the closest one with a free plan.

Reviews

No reviews yet

Listing history

  1. Nov 14, 2026Checked against shotoku.dev
  2. Nov 14, 2026Listed