EveryDev.ai
Subscribe
Home
Tools

5,901+ AI tools

  • Featured
  • Gems
  • Trending
  • New
  • Rate tools
  • Compare
  • Arena
Categories
  • Agents2972
  • Coding2248
  • Infrastructure1067
  • Projects818
  • MCP794
  • Research744
  • Analytics565
  • Marketing527
  • Design488
  • Testing476
  • Data442
  • Security412
  • Integration288
  • Communication284
  • Prompts283
  • Extensions238
  • Voice213
  • DevOps190
  • Commerce183
  • Web128
  • Learning123
  • Finance44
AI Tools by Topic
  • AI Coding Assistants
  • Agent Frameworks
  • MCP Servers
  • AI Prompt Tools
  • Vibe Coding Tools
  • AI Design Tools
  • AI Database Tools
  • AI Website Builders
  • AI Testing Tools
  • LLM Evaluations
Follow Us
  • X / Twitter
  • LinkedIn
  • Reddit
  • Discord
  • Threads
  • Bluesky
  • Mastodon
  • YouTube
  • GitHub
  • Instagram
Get Started
  • Users
  • Rate Tools
  • About
  • Editorial Standards
  • Corrections & Disclosures
  • Community Guidelines
  • Advertise
  • Contact Us
  • Newsletter
  • Submit a Tool
  • Start a Discussion
  • Write A Blog
  • Share A Build
  • Terms of Service
  • Privacy Policy
Explore with AI
  • ChatGPT
  • Gemini
  • Claude
  • Grok
  • Perplexity
Agent Experience
  • llms.txt
Theme
With AI, Everyone is a Dev. EveryDev.ai ©
    1. Home
    2. Tools
    3. Shotoku
    Shotoku icon

    Shotoku

    MCP Tools
    Featured

    Local-first spend controls for AI agents that enforce budgets, human approvals, and a tamper-evident audit trail before a payment happens.

    Visit Website

    At a Glance

    Pricing
    Open Source
    Free tier available

    Everything you need to authorize, approve, and audit agent actions on your machine. MIT-licensed core.

    Enterprise: Custom/contact

    Engagement

    Available On

    Windows
    macOS
    Linux
    Web
    API

    Resources

    WebsiteDocsGitHubllms.txt

    Topics

    MCP ToolsAccess ControlHuman-in-the-Loop Training

    Alternatives

    WebMCPPreloopwithHuman
    Developer
    ShotokuEst. 2026

    Listed Nov 2026

    About Shotoku

    Shotoku is an open-source, MIT-licensed authorization layer for AI agent spending, built by Julius and Issa, two developers from France. It evaluates each agent action against a YAML policy on the user's own machine, then records the decision in a local ledger. The project describes itself as beta, with a stable, tested core authorization path.

    What It Is

    Shotoku sits between an AI agent and the thing it wants to do, such as a purchase, a paid API call, an MCP tool call, or code execution. Each request returns a structured decision: approved, denied, or pending human approval, with reasons and a decision ID. It holds no funds, stores no credentials or private keys, and never signs or settles a payment; the agent's own wallet or payment rail does the executing. It runs without an account, cloud, or telemetry.

    How Policies and Approvals Work

    A policy.yaml file defines rules per resource, with per-transaction limits (maxAmount), rolling 24-hour limits (maxDailyAmount), and per-agent daily budgets. A rule can use overLimit: pending_approval so small amounts pass and large ones wait for a human. If the policy is missing or invalid, the ledger cannot be trusted, or an amount is unknown, the agent is not approved. A human approval lets through one identical retry within 15 minutes, and agents cannot approve their own decisions.

    Pending decisions can be answered from the terminal, a TUI, a local dashboard (shotoku ui), or Telegram buttons on a phone.

    MCP Wrapping and Payment Rails

    Shotoku ships an MCP server with two modes. In wrapping mode it sits in front of the MCP servers an agent already uses, exposing each tool as server__tool and deciding every call before forwarding it. Money-moving tools are priced, and PayPal's MCP server is priced out of the box, including looking up a capture so a refund with no amount is priced at the full amount. In advisory mode the agent calls authorize_action itself. The README also describes x402 payment authorization after the price is known and before the wallet signs.

    Audit and Verification

    Every decision and human answer is written to a local, append-only, hash-chained JSONL ledger at data/decisions.jsonl. Editing a past record breaks the chain on later reads. Signed snapshots of the policy and ledger head can be created and verified, and approved decisions can carry short-lived HMAC-signed receipts that a payment proxy or CI job verifies before executing.

    Setup Path

    Install the CLI with npm install -g shotoku-cli, run shotoku demo for four simulated decisions, then shotoku init and shotoku authorize to evaluate a spend. A TypeScript function, authorize, is available from @shotoku/core. The README's own comparison, run on one model against an offline PayPal stand-in, reports that an unguarded agent refunded a $480 item in 5 of 5 runs while the same agent behind Shotoku was held for a human each time; the authors call it a small sample.

    Shotoku - 1

    Community Discussions

    Be the first to start a conversation about Shotoku

    Share your experience with Shotoku, ask questions, or help others learn from your insights.

    Pricing

    OPEN SOURCE

    Open Source

    Everything you need to authorize, approve, and audit agent actions on your machine. MIT-licensed core.

    • authorize() core and policy engine
    • Hash-chained decision ledger
    • CLI and terminal UI
    • MCP server for Claude, Cursor, and more
    • x402 payment authorization

    Enterprise

    Governance, enforcement, and compliance for organizations where agents act at scale.

    Custom
    contact sales
    • Signed decision receipts and enforcement integration
    • RBAC on approvals
    • Compliance export packs (EU AI Act, SOC 2)
    • SSO / SAML / SCIM
    • Support SLA
    View official pricing

    Capabilities

    Key Features

    • Per-transaction and rolling 24-hour spend limits per vendor
    • Per-agent daily budgets across all vendors
    • Human approval queue with pending_approval verdicts and overLimit rules
    • Approve or deny via terminal, TUI, local dashboard, or Telegram
    • MCP server with wrapping mode that gates every tool call
    • Built-in pricing for PayPal's MCP server
    • x402 payment authorization before the wallet signs
    • Append-only hash-chained local audit ledger
    • Signed snapshots and HMAC-signed receipts
    • Fails closed when policy or ledger is invalid
    • YAML policy configuration
    • Local-first with no account, cloud, or telemetry

    Integrations

    PayPal
    x402
    MCP
    Claude
    Telegram
    TypeScript
    AG Grid
    API Available
    View Docs

    Demo Video

    Shotoku Demo Video
    Watch on YouTube

    Ratings & Reviews

    No ratings yet

    Be the first to rate Shotoku and help others make informed decisions.

    Rate other tools you’ve used

    Developer

    Shotoku Team

    Shotoku is built by Julius and Issa, two developers from France. They publish the project as open source under the MIT license. They say they plan to grow it into an enterprise product and welcome contributions and feedback.

    Founded 2026

    Used by

    PayPal is the featured integration and…
    Read more about Shotoku Team
    WebsiteGitHubX / Twitter
    1 tool in directory

    Similar Tools

    WebMCP icon

    WebMCP

    A proposed JavaScript browser API that lets web developers expose app functionality as tools invokable by AI agents and assistive technologies for human-in-the-loop workflows.

    Preloop icon

    Preloop

    Human-in-the-loop approval platform for AI agents with MCP proxy, tool management, and team-based workflows.

    withHuman icon

    withHuman

    Approval layer for AI agents that holds consequential tool calls for rule-based or human approval before they run.

    Browse all tools

    Related Topics

    MCP Tools

    Tools built with the Model Context Protocol for specific tasks.

    172 tools

    Access Control

    AI-enhanced tools for managing authentication and authorization.

    66 tools

    Human-in-the-Loop Training

    Platforms that connect organizations with vetted human experts to annotate, label, evaluate, and align AI models, ensuring high-quality training datasets and accurate model evaluation through human judgment.

    73 tools
    Browse all topics
    Back to all toolsSuggest an edit
    ratings
    discussions