K3 Blockchain Agent Skill

The K3 Blockchain Agent is a skill for Claude Code that turns your AI coding assistant into a full blockchain workflow builder. Describe what you want to automate in plain language — monitoring a wallet, tracking a DeFi pool, sending daily crypto reports — and Claude will build, deploy, and verify the workflow for you on K3.

No workflow builder UI required. No manual node configuration. Just describe what you want, and Claude handles the rest.

---

What You Can Build

Any combination of data source, AI analysis, and delivery channel:

Monitoring & Alerts

• Alert me on Telegram when this wallet moves more than 10 ETH

• Notify me on Slack when a large swap happens on this Uniswap pool

• Watch this smart contract and alert me when a specific event fires

Scheduled Reports

• Send me a daily email with a summary of the top Uniswap V3 pools by volume

• Every morning, pull the latest metrics for my DeFi portfolio and email me a report

• Weekly recap of what's happening with this token — TVL, volume, price action

Conversational Bots

• Build a Telegram bot I can chat with to check wallet balances

• Create a bot that answers questions about any token's market data

• Telegram assistant that lets me query on-chain data by sending a message

Data Collection

• Track this pool's TVL daily and store the data in Google Sheets

• Collect hourly price data for these tokens and save to my database

Automated Trading

• Swap tokens on Uniswap when the price hits a target

• Rebalance my portfolio on a weekly schedule

• Transfer tokens automatically when a condition is met

---

Prerequisites

1. K3 MCP must be connected

The skill requires the K3 MCP server to be active. If you haven't set it up yet, follow the K3 MCP Server setup guide first — the steps vary by client.

2. Skill must be installed

See the Installation section below for your client.

---

Installation

Claude Code

Copy the skill folder into your project's .claude/skills/ directory:

mkdir -p .claude/skills
cp -r /path/to/k3-blockchain-agent .claude/skills/

The skill is automatically loaded when you open a Claude Code session in that project. No additional configuration is needed. Verify it's active by running /mcp to confirm the K3 server is connected.

---

OpenClaw

Copy the skill folder into your OpenClaw skills directory:

# Global (available to all agents)
mkdir -p ~/.openclaw/skills
cp -r /path/to/k3-blockchain-agent ~/.openclaw/skills/

# Or project-level (available in a specific workspace)
mkdir -p ./skills
cp -r /path/to/k3-blockchain-agent ./skills/

Then restart the Gateway so OpenClaw picks up the new skill:

openclaw gateway restart

Verify the skill is loaded:

openclaw skills list --eligible | grep k3

You should see k3-blockchain-agent in the output. The skill is now available to any agent in your OpenClaw config that has the K3 MCP connected.

---

How to Use It

Once the skill and MCP are active, just describe what you want in plain language. Your assistant will ask any clarifying questions it needs, then build and deploy the workflow.

Example prompts to get started:

Send me a daily Telegram message with the top 5 Uniswap V3 pools by volume on Ethereum

Monitor wallet 0x... on Arbitrum and alert me on Telegram whenever it moves more than 5 ETH

Build a Telegram bot I can chat with to check token prices and wallet balances

Every morning, email me a report on the WETH/USDC pool — TVL, volume, fees, and any notable changes

Alert me on Slack if any swap over $500k happens on this pool: 0x...

Your assistant will work through the request with you, confirm the details, and then build the workflow. Expect a short back-and-forth — 2–4 rounds is normal before the workflow is ready to deploy.

Using with OpenClaw

With OpenClaw, you interact through your connected messaging app (Telegram, Discord, WhatsApp, etc.) rather than a terminal. Send the same natural language requests directly to your OpenClaw bot:

"Set up a daily DeFi report for me — WETH/USDC pool on Uniswap V3, email me every morning"

OpenClaw will use the K3 skill to build and deploy the workflow, then confirm when it's live. You can manage and iterate on workflows entirely from within the chat, without opening a terminal.

---

What Happens Under the Hood

Your assistant uses the K3 MCP tools to orchestrate everything:

1. Understands your request — identifies the data source, schedule, analysis, and delivery channel

2. Finds the right data — checks your team's connected integrations, searches for relevant APIs or subgraphs, and tests the data before building

3. Builds the workflow — calls generateWorkflow with a detailed prompt describing exactly what to build

4. Reviews and refines — iterates with the orchestrator via editGeneratedWorkflow until the config is correct

5. Deploys and verifies — deploys the workflow, runs a manual test, and confirms the full chain worked

---

Workflow Building Blocks

Understanding what's available helps you ask for more specific things.

Triggers — what starts the workflow

Trigger	When to use
Scheduled	Daily digests, hourly monitors, periodic reports
Smart Contract Event	React to swaps, transfers, mints, liquidations, governance votes
Wallet Activity	Whale watching, portfolio alerts, incoming payment detection
Telegram Chatbot	Conversational AI bot — responds to messages from users
On-Demand / Manual	Ad-hoc queries, tools you run when you want

Data Sources — where the workflow gets its data

Source	What it fetches
Read Market Data	Token prices, volume, market cap
Read Wallet	Balances, token holdings, transaction history
Read Smart Contract	On-chain contract state, pool reserves, interest rates
Read Graph	DeFi metrics from TheGraph subgraphs (TVL, volume, fees)
Read NFT	Collection data, floor prices, traits, holders
Read API	Any REST or GraphQL API
AI Web Scraper	Structured data extracted from any web page
AI Agent with tools	AI that decides what to fetch and calls multiple sources

AI Functions — how the workflow thinks

Function	What it does
AI Service	Summarizes, analyzes, or formats data
AI Agent	Autonomous reasoning with tool access for complex tasks
AI Conditional	Branches workflow logic based on AI judgment

Delivery & Actions — what the workflow does with results

Output	Use case
Email	Reports, daily digests, alert emails
Telegram	Instant alerts, bot replies, real-time updates
Slack	Team notifications, shared monitoring
Google Sheets / API	Data collection, dashboards, external systems
Uniswap / Coinbase / Kyberswap	Automated trading and swaps
Token Transfer	Automated payments and treasury management
Write Smart Contract	On-chain actions triggered by workflow logic

---

Example Workflows

Daily DeFi Pool Report

"Send me a daily email with metrics for the WETH/USDC pool on Uniswap V3"

[Daily schedule] → [Read Graph: query Uniswap V3 subgraph] → [AI Service: summarize TVL, volume, fees, highlight changes] → [Email]

---

Whale Wallet Alert

"Alert me on Telegram when wallet 0x... moves more than 10 ETH"

[Wallet Activity trigger] → [Read Wallet: get transaction details] → [AI Conditional: check if amount > 10 ETH] → [Telegram alert]

---

Telegram On-Chain Assistant

"Build a Telegram bot I can ask about token prices and wallet balances"

[Telegram Chatbot trigger] → [AI Agent with tools: CoinGecko + Read Wallet] → [Telegram reply]

---

Multi-Protocol Yield Comparison

"Every Monday, compare yields across Uniswap, Aave, and Curve and email me the best opportunities"

[Weekly schedule] → [Read Graph: Uniswap pools] → [Read API: Aave rates] → [Read API: Curve pools] → [AI Service: compare and rank] → [Email]

---

Tips for Best Results

Be specific about delivery. Tell your assistant exactly where you want results — email address, Telegram bot name, Slack channel. It won't guess.

Name the protocol and chain. "Uniswap V3 on Arbitrum" is better than "Uniswap". Pool addresses and contract addresses are chain-specific — include them if you have them.

Describe the analysis you want. "Summarize the pool" is vague. "Flag any TVL change over 5% and compare volume to the 7-day average" gives the assistant something concrete to work with.

For trading workflows, review carefully. Automated trading involves real funds. Your assistant will explain exactly what conditions trigger what actions before deploying — read this carefully.

Iterate freely. The workflow builder is conversational. If the first version isn't quite right, just say what to change. There's no need to start over.

---

Troubleshooting

"K3 MCP isn't connected" (Claude Code)

Run /mcp to check which servers are connected. If K3 isn't listed, follow the K3 MCP Server setup guide.

"K3 MCP isn't connected" (OpenClaw)

Check that the K3 MCP is configured in your ~/.openclaw/openclaw.json under the mcpServers block for the relevant agent, and that you've run openclaw gateway restart. See the K3 MCP Server setup guide for the full config.

Skill not found in OpenClaw

Run openclaw skills list --eligible | grep k3. If it doesn't appear, confirm the skill folder is in ~/.openclaw/skills/ or <workspace>/skills/ and that it contains a SKILL.md file. Restart the gateway after adding it.

Workflow deployed but data is empty

The most common cause is a wrong API endpoint, subgraph URL, or contract address. These are chain-specific — the same pool has a different address on Ethereum vs. Arbitrum. Tell your assistant what you expected to see and it will investigate and fix.

AI analysis is too generic

Add more detail to your request. Instead of "analyze the pool", say "highlight any TVL change over 5%, compare this week's volume to last week, and flag any unusual fee spikes." The more specific your intent, the better the analysis.

Telegram or email notification didn't arrive

Check that the integration is connected in K3 (Integrations in the left sidebar). For Telegram, also verify the bot token and chat ID. For email, check your spam folder.

Scheduled workflow stopped running

Ask your assistant to check the workflow status — it will call getWorkflowById and getWorkflowRuns to diagnose whether it's paused or erroring.