autonomous-agent – OpenClaw Skill

---

name: autonomous-agent description: CreditNexus x402 agent. Use when the user wants to create and use wallets (Aptos/EVM), fund wallets, check balances, run stock predictions, backtests, open bank accounts, or transfer/swap tokens. Payment-protected MCP tools with x402 flow (Aptos + Base). Agent handles 402 → pay → retry autonomously. metadata: {"clawdbot":{"emoji":"📈","homepage":"https://github.com/FinTechTonic/autonomous-agent","requires":{"bins":["node","npm"]}}}

Autonomous Agent (x402) – Skill

Autonomous agent that creates, funds, and uses Aptos and EVM wallets, then calls x402-paid MCP tools (stock prediction, backtest, open bank account). Handles payment flow (402 → pay → retry) with Aptos and Base. Use this skill when the user wants wallet setup, funding, balances, predictions, backtests, banking, or token operations.

---

Tasks you can do

Wallets – create and manage

• Create Aptos wallet – create_aptos_wallet (optionally network: "testnet" or "mainnet"). Agent can have multiple Aptos wallets.
• Create EVM wallet – create_evm_wallet (optionally network: "testnet" or "mainnet"). Agent can have multiple EVM wallets.
• List wallet addresses – get_wallet_addresses returns all Aptos and EVM addresses (with network tags) for whitelisting and funding.

Fund wallets

• Fund Aptos wallet – credit_aptos_wallet: on devnet uses programmatic faucet; on testnet returns instructions and Aptos faucet URL. Needed for run_prediction and run_backtest (~6¢ USDC).
• Fund EVM wallet – fund_evm_wallet: returns address and instructions (Base Sepolia faucet, etc.). Needed for open_bank_account (~$3.65 on Base).

User must whitelist every agent address at the onboarding flow (e.g. http://localhost:4024/flow.html or your MCP server’s flow) before paid tools succeed.

Check balances

• Aptos balance – balance_aptos (USDC for the agent wallet).
• EVM balance – balance_evm (native token on a chain: base, baseSepolia, ethereum, etc.).

Paid MCP tools (x402)

• Stock prediction – run_prediction (symbol, horizon in days). Cost ~6¢ (Aptos).
• Backtest – run_backtest (trading strategy). Cost ~6¢ (Aptos).
• Open bank account – open_bank_account (CornerStone bank link). Cost ~$3.65 (EVM/Base).

The agent handles 402 Payment Required automatically: verify → settle → retry with payment signature.

CLI (from repo root)

Task	Command
Generate Aptos wallet	node src/setup-aptos.js
Generate EVM wallet	node src/setup.js
Show addresses for whitelist	node src/show-agent-addresses.js
Credit Aptos (devnet)	npm run credit:aptos (set APTOS_FAUCET_NETWORK=devnet)
EVM balance	node src/balance.js <chain>
Transfer ETH/tokens	node src/transfer.js <chain> <to> <amount> [tokenAddress]
Swap tokens (Odos)	node src/swap.js <chain> <fromToken> <toToken> <amount>
Run agent	node src/run-agent.js "Run a 30-day prediction for AAPL" or node src/run-agent.js (interactive)

---

When to use this skill

Use when the user wants to:

• Create or use Aptos or EVM wallets (testnet or mainnet).
• Fund agent wallets (faucet instructions or programmatic credit).
• Check Aptos or EVM balances.
• Run stock predictions or backtests (paid via Aptos).
• Open a bank account (paid via Base).
• Transfer or swap tokens from the agent wallet (via CLI or context).

---

Setup

1. Install: From repo root: npm install. Copy .env.example to .env.
2. Configure: Set MCP_SERVER_URL, X402_FACILITATOR_URL, HUGGINGFACE_API_KEY (or HF_TOKEN), LLM_MODEL, and wallet paths (APTOS_WALLET_PATH, EVM_WALLET_PATH or EVM_PRIVATE_KEY).
3. Wallets: Create via agent tools (create_aptos_wallet, create_evm_wallet) or CLI (node src/setup-aptos.js, node src/setup.js). Fund and whitelist all addresses at the MCP server’s flow (e.g. /flow.html).

---

Run the agent

From the repository root (where package.json and src/ live):

node src/run-agent.js "Create an Aptos wallet, then run a 30-day prediction for AAPL"
# Or interactive
node src/run-agent.js

Source: FinTechTonic/autonomous-agent

Autonomous Agent – x402 MCP + LangChain.js

Autonomous AI agent for x402-paid MCP tools: predict tickers, backtest strategies, open bank accounts. Pays with Aptos or EVM via the x402 facilitator—no OpenAI key; uses Hugging Face for the LLM. Built for Cursor, OpenClaw/Moltbot, and headless runs.

Why?

Most agent demos need a full backend and separate API keys. This agent talks to an x402 MCP server, pays with its own Aptos and EVM wallets (verify → settle), and uses an OpenAI-compatible LLM (e.g. Hugging Face). You get run_prediction, run_backtest, and open_bank_account with minimal setup—whitelist your agent, fund wallets, run.

Install

From npm (recommended)

npm install cornerstone-autonomous-agent

Copy the package’s .env.example into your project and set MCP_SERVER_URL, X402_FACILITATOR_URL, HUGGINGFACE_API_KEY, LLM_MODEL, and wallet paths. See Config.

From source

git clone https://github.com/FinTechTonic/autonomous-agent.git && cd autonomous-agent
npm install

OpenClaw / MoltBook / Moltbot: Load the skill from skills/autonomous-agent/ or adapters/openclaw/SKILL.md. See MoltBook / OpenClaw below. Then point MCP_SERVER_URL at your MCP server.

Quick Start

# Generate Aptos wallet (for prediction/backtest)
node src/setup-aptos.js

# Generate EVM wallet (for open_bank_account)
node src/setup.js

# Whitelist your agent at the onboarding flow (e.g. https://borrower.replit.app/flow.html)
node src/show-agent-addresses.js

# Run the agent (demo: balance + AAPL prediction)
npx cornerstone-agent "Run a 30-day prediction for AAPL"
# or from source:
node src/run-agent.js "Run a 30-day prediction for AAPL"

Config

Copy .env.example to .env and set:

Variable	Description
MCP_SERVER_URL	x402 MCP server base URL (e.g. https://borrower.replit.app or http://localhost:4023)
X402_FACILITATOR_URL	Facilitator base URL for Aptos (verify/settle). Use public (e.g. https://x402-navy.vercel.app/facilitator) for open_bank_account.
X402_EVM_FACILITATOR_URL	Optional. EVM facilitator; defaults to X402_FACILITATOR_URL. Set to public when using local Aptos facilitator.
LLM_BASE_URL	OpenAI-compatible base URL (default https://router.huggingface.co/v1)
HUGGINGFACE_API_KEY or HF_TOKEN	Hugging Face API key
LLM_MODEL	Model ID (e.g. meta-llama/Llama-3.2-3B-Instruct)
APTOS_WALLET_PATH	Aptos wallet JSON path. Multi-wallet: ~/.aptos-agent-wallets.json.
EVM_WALLET_PATH	EVM wallet path. Multi-wallet: ~/.evm-wallets.json. Or set EVM_PRIVATE_KEY.
BASE_SEPOLIA_RPC	Optional; Base Sepolia RPC for open_bank_account

Commands

Command	Description
node src/setup.js	Generate EVM wallet (single; for multi use agent tool create_evm_wallet)
node src/setup-aptos.js	Generate Aptos wallet (single; for multi use create_aptos_wallet)
node src/show-agent-addresses.js	Print all Aptos and EVM addresses for whitelisting at flow.html
npm run credit:aptos	Credit Aptos agent (devnet: programmatic; testnet: instructions)
node src/balance.js <chain>	EVM balance
node src/run-agent.js [message]	Run agent (or npx cornerstone-agent [message] when installed from npm)
npm run agent	Same as run-agent.js

Crediting Aptos: Testnet has no programmatic faucet—use Aptos testnet faucet. Devnet: APTOS_FAUCET_NETWORK=devnet npm run credit:aptos. See Canteen – Aptos x402.

MCP Tools

Tool	Description	Cost
run_prediction	Stock prediction (symbol, horizon, strategy)	~6¢ (Aptos)
run_backtest	Backtest trading strategy	~6¢ (Aptos)
open_bank_account	CornerStone bank link / open bank account	~$3.65 (Ethereum/Base)

Supported Networks

Network	Use	Cost
Aptos testnet	run_prediction, run_backtest	~6¢ USDC
Base Sepolia	open_bank_account (testnet)	~$3.65
Base (mainnet)	open_bank_account (production)	~$3.65

x402 Flow

Agent calls MCP tool
  → Server returns 402 + payment requirements
  → Agent builds payment (Aptos or EVM), calls facilitator /verify then /settle
  → Agent retries request with PAYMENT-SIGNATURE header
  → Server returns result + request_payload, response_payload, payment_receipt

Architecture

autonomous/
├── src/
│   ├── agent/           # ReAct agent, LLM, tool wiring
│   ├── lib/
│   │   ├── mcp/          # MCP client + 402 handle/retry
│   │   ├── aptos/        # Aptos wallet, balance, signPayment
│   │   ├── evm/          # EVM wallet, signPayment (Base)
│   │   └── x402/         # Payment types, verify/settle flow
│   ├── run-agent.js      # Entrypoint
│   ├── setup.js          # EVM wallet generation
│   ├── setup-aptos.js    # Aptos wallet generation
│   └── show-agent-addresses.js
├── skills/               # MoltBook/OpenClaw (AgentSkills layout)
│   └── autonomous-agent/
│       └── SKILL.md
├── adapters/             # OpenClaw, OpenAI, Anthropic, local
├── .env.example
└── package.json

Core pieces: lib/mcp — MCP client and 402 retry; lib/aptos / lib/evm — wallets and payment signing; lib/x402 — verify/settle; agent/ — LangChain.js ReAct agent and tools.

Tech Stack

• Runtime: Node.js 18+
• Agent: LangChain.js (ReAct), OpenAI-compatible LLM (e.g. Hugging Face)
• MCP: Model Context Protocol + x402 payment flow
• Chains: Aptos (viem-style + @aptos-labs/ts-sdk), EVM (viem) for Base Sepolia/Base
• Payments: x402 facilitator (verify/settle), local wallet storage

Security

• Wallets: Stored locally (e.g. ~/.aptos-agent-wallets.json, ~/.evm-wallets.json); private keys not logged or sent except as signed payloads to the facilitator.
• Payments: Only verify/settle go to the facilitator; no custody of funds by the MCP server.
• Whitelist: Agent addresses must be allowlisted at the onboarding flow before paid tools succeed.

Capability + adapters

• Capability: Core (src/) — MCP client, x402 flow, local tools. No OpenAI/Claw/Anthropic logic in code.
• Adapters: How each platform uses the capability:
  • MoltBook / OpenClaw / Moltbot: skills/autonomous-agent/SKILL.md (AgentSkills-compatible; see below) or adapters/openclaw/SKILL.md
  • OpenAI: adapters/openai/openapi.yaml — Custom GPTs / Assistants
  • Claude / Anthropic: adapters/anthropic/tools.json — Claude tools
  • Local / OSS: adapters/local/README.md — LM Studio, AutoGen, CrewAI

MoltBook / OpenClaw

This repo is optimized for easy skill loading in MoltBook and OpenClaw (Claude, Anthropic, OpenAI, and other providers work via the same agent; the skill tells the assistant how to run it).

• Skill layout: The skill lives in skills/autonomous-agent/SKILL.md (AgentSkills-compatible, single-line frontmatter, metadata.openclaw gating). OpenClaw/MoltBook loads skills from skills/ subfolders.
• Load options:
  1. extraDirs: Add this repo path to ~/.openclaw/openclaw.json under skills.load.extraDirs. OpenClaw will scan skills/ and load autonomous-agent.
  2. Workspace: Clone the repo and use it as your OpenClaw workspace; workspace skills are loaded from <workspace>/skills.
  3. Managed skills: Copy skills/autonomous-agent to ~/.openclaw/skills/ for all agents on the machine.
  4. ClawHub: When published, install with clawhub install autonomous-agent (installs into ./skills by default).
• Config: In skills.entries["autonomous-agent"] you can set enabled, env, or apiKey (maps to primaryEnv). Ensure MCP_SERVER_URL, x402 facilitator URLs, and LLM/env are set for the agent run.
• Run: From the repository root (parent of skills/), run node src/run-agent.js "your message".

Example ~/.openclaw/openclaw.json to load this repo’s skills:

{
  "skills": {
    "load": {
      "extraDirs": ["/path/to/autonomous-agent"]
    },
    "entries": {
      "autonomous-agent": {
        "enabled": true,
        "env": { "MCP_SERVER_URL": "https://borrower.replit.app" }
      }
    }
  }
}

Publishing this skill on MoltBook / OpenClaw (ClawHub)

ClawHub is the public skill registry for MoltBook and OpenClaw. Publishing lets others install your skill with clawhub install autonomous-agent.

1. Install the ClawHub CLI

   npm i -g clawhub
   

   or pnpm add -g clawhub.

2. Log in (GitHub account required; must be at least one week old to publish)

   clawhub login
   

   This opens a browser to authenticate. For CI or headless: clawhub login --token <token>.

3. Publish the skill from the repo root:

   clawhub publish ./skills/autonomous-agent --slug autonomous-agent --name "Autonomous Agent (x402)" --version 1.0.0 --changelog "Initial release" --tags latest
   

   • --slug must match the folder name (autonomous-agent) so installs work as clawhub install autonomous-agent.
   • Omit --version / --changelog to be prompted. For later releases, bump version (e.g. --version 1.0.1) and add a --changelog line.

4. Update an existing skill (after you change SKILL.md or metadata): run publish again with a new version:

   clawhub publish ./skills/autonomous-agent --slug autonomous-agent --name "Autonomous Agent (x402)" --version 1.0.1 --changelog "Description of changes" --tags latest
   

   Alternatively, clawhub sync --root . --all --bump patch --changelog "Description" scans the repo and publishes changed skills with a bumped version.

After publishing, the skill appears on clawhub.ai and users can run clawhub install autonomous-agent (into ./skills by default).

Deployment order

1. x402 facilitator — Use public (e.g. https://x402-navy.vercel.app/facilitator) for full demo; or run local and set X402_EVM_FACILITATOR_URL to public for open_bank_account.
2. MCP server — x402-enabled (e.g. https://borrower.replit.app or run locally).
3. Agent — node src/run-agent.js or PM2 (pm2 start ecosystem.config.cjs --only agent-autonomous from repo root).

References

• Source: FinTechTonic/autonomous-agent
• Canteen App – Aptos x402 — wallet hydration and crediting
• LangChain.js MCP
• Hugging Face Inference – OpenAI-compatible

License

GPL-2.0-only. Use of this software is also subject to the Responsible AI License (RAIL). See LICENSE.md (GPL-2) and RAIL.