Intent-based payments for OpenClaw agents.

Prepaid wallet. Payment tasks. One hidden execution card. Account inbox.

Website · Docs Site · OpenClaw Quickstart · Docs · Skill

This is the public MoneyClaw repository.

It is the place for:

• public product docs
• the MoneyClaw OpenClaw skill
• examples and starter prompts
• high-level architecture and safety model

It is not the full internal product monorepo.

Internal infrastructure, production runbooks, incidents, security hardening notes, deployment details, and provider-specific operational playbooks are intentionally not published here.

MoneyClaw gives OpenClaw agents a prepaid wallet and an auditable payment-execution layer for online purchases and recurring subscriptions.

With MoneyClaw, an agent can:

• check wallet and inbox readiness
• create payment tasks
• start recurring payment setups
• prepare or reuse one hidden execution card under the account when execution needs it
• complete browser checkout
• inspect account inbox messages when needed
• verify the final transaction result

Execution note:

• cards are internal execution artifacts, not the main public product object
• public integrations should lead with wallet, payment tasks, subscriptions, inbox, and merchant flows

Secondary workflows include:

• hosted invoices
• merchant checkout links
• merchant and acquiring flows when explicitly requested

Install the skill:

clawhub install moneyclaw

Then:

1. Create a MoneyClaw API key.
2. Store it in OpenClaw as MONEYCLAW_API_KEY.
3. Start a fresh OpenClaw session.
4. Ask the agent to inspect the account before its first purchase.

Starter prompt:

Check my MoneyClaw account and tell me if the wallet, inbox, and payment tasks are ready.

MoneyClaw is built around bounded, inspectable agent payments.

• prepaid balances keep risk capped
• payment tasks and subscriptions hold approval and audit state
• one hidden execution card stays under the account and is reused across payment tasks and subscriptions
• receipts and account messages stay visible in a dedicated inbox
• the user can keep dashboard approval on or deliberately enable account-level agent auto-approval

Important boundary:

MoneyClaw is for user-authorized payments. It is not designed to bypass issuer, merchant, fraud, KYC, sanctions, geography, or verification controls.

Read the full model in docs/security-model.md.

Start here:

1. docs/getting-started.md
2. docs/openclaw-quickstart.md
3. docs/api-overview.md
4. moneyclaw-skill/SKILL.md

assets/            branding and repo visuals
docs/              public documentation
examples/          prompts and curl examples
moneyclaw-skill/   public OpenClaw skill files

Published here:

• what MoneyClaw is
• how to use it
• how the public skill is structured
• how the public API is meant to be understood
• how the safety and trust model works

Not published here:

• production IPs and infra
• internal incidents and postmortems
• operational security notes
• private merchant-specific tactics
• internal team process and private agent instructions

pnpm install
cp .env.example .env
docker compose up -d
export REDIS_URL=redis://localhost:6379

import { getOrSetJson, buildCacheKey } from "moneyclaw-public/redis/cache";

const task = await getOrSetJson(
  buildCacheKey("payment-task", ["task-123"]),
  300,
  async () => fetchFromApi(),
);

Helpers in src/redis/ support caching, webhook dedup, and distributed locks for local tooling. When REDIS_URL is unset, they fail open.

The public OpenClaw skill talks to the MoneyClaw API directly — Redis is optional for skill authors.

• Product site: moneyclaw.ai
• Docs hub: elvismusli.github.io/moneyclaw
• Security contact: SECURITY.md

Code and docs in this repository are available under the MIT License, unless noted otherwise.