Pond3r Skill - Query Onchain Data

Pond3r Crypto Intelligence Skill

Use this skill when the agent needs to query crypto/DeFi data: yields, protocol metrics, token opportunities, market analysis, or blockchain analytics. Pond3r provides an MCP server with read-only SQL access to curated datasets.

Prerequisites

Setup: Where Is the Agent Running?

The agent needs MCP tools (list_datasets, get_schema, query) to use Pond3r. Those tools come from the runtime that executes the agent — not from the skill or env vars alone.

| Runtime | How to enable Pond3r | |---------|------------------------| | Cursor | Cursor Settings → MCP Servers → Add server (URL + Authorization header). See MCP Connection below. | | Claude Desktop | Add Pond3r to claude_desktop_config.json under mcpServers. Restart Claude. | | Claude Code | Run claude mcp add pond3r-data ... (see below). | | OpenClaw (Docker/Telegram) | Use the CLI scripts below. They call Pond3r MCP via HTTP. Set POND3R_API_KEY in .env; the agent runs the scripts and parses JSON output. |

No extra info for the agent — the skill is enough. For runtimes with native MCP, add the API key in the MCP server config. For OpenClaw, use the scripts and POND3R_API_KEY in env.

MCP Connection

| Setting | Value | |---------|-------| | URL | https://mcp.pond3r.xyz/mcp | | Transport | Streamable HTTP | | Auth | Authorization: Bearer |

Cursor

1. Open Cursor → Settings (⌘+,) → MCP 2. Click Add new MCP server 3. Configure: - URL: https://mcp.pond3r.xyz/mcp - Headers: Authorization: Bearer - Replace with your key from makeit.pond3r.xyz/api-keys - Some clients support Authorization: Bearer ${POND3R_API_KEY} if that env var is set 4. Save and restart Cursor so tools load 5. Verify: start a new chat and ask for stablecoin yields — the agent should call list_datasets, get_schema, query

Claude Code

claude mcp add pond3r-data \
  --transport http \
  https://mcp.pond3r.xyz/mcp \
  --header "Authorization: Bearer "

Claude Desktop (claude_desktop_config.json)

{
  "mcpServers": {
    "pond3r": {
      "type": "http",
      "url": "https://mcp.pond3r.xyz/mcp",
      "headers": {
        "Authorization": "Bearer "
      }
    }
  }
}

CLI Scripts (OpenClaw / Any Runtime)

When MCP tools are not available (e.g. OpenClaw/Telegram), use these scripts. They call Pond3r MCP over HTTP. Requires POND3R_API_KEY in env (e.g. in .env loaded by docker-compose).

Scripts live at /opt/pond3r-skill-scripts/ in the Docker image. When running locally, use ceo-agent/skills/pond3r-skill/scripts/ or the workspace-relative path.

1) List datasets

node /opt/pond3r-skill-scripts/list-datasets.mjs

Output: JSON with all datasets and tables.

2) Get schema for a dataset

node /opt/pond3r-skill-scripts/get-schema.mjs --dataset-id 

3) Run a SQL query

node /opt/pond3r-skill-scripts/query.mjs --dataset-id  --sql "SELECT * FROM stablecoin_yields LIMIT 10"

Or from file:

node /opt/pond3r-skill-scripts/query.mjs --sql-file /tmp/query.sql

Script workflow

1. Run list-datasets.mjs to discover datasets and table names. 2. Run get-schema.mjs --dataset-id to see columns and types. 3. Run query.mjs --dataset-id --sql "SELECT ..." with valid SQL (SELECT only, bare table names, LIMIT where appropriate). 4. Parse the JSON output and summarize for the user.

Failure handling

Available Tools (Native MCP)

| Tool | Purpose | |------|---------| | list_datasets | List all datasets and their tables | | get_schema | Get column names, types, descriptions for a dataset | | query | Execute read-only SQL against a dataset |

Query Rules

Use Cases

1. Protocol Intelligence - Track AI agent launches, token graduations, protocol metrics - Daily yield farming reports across Aave, Compound, Convex

2. Market Opportunity Detection - New tokens on Uniswap with rising liquidity - Tokens with <$500K market cap and rising liquidity - Polymarket trades with highest volume

3. Risk-Adjusted Analysis - Multi-dimensional risk scoring (volatility, liquidity, market structure) - Liquidation risk monitoring for DeFi positions - Whale activity tracking

4. Cross-Protocol & Cross-Chain - Compare USDC yields across Aave and Compound on Arbitrum - Bridge volume analysis, ecosystem health comparison - Arbitrage opportunity detection

5. Structured Data for Decisions - Statistical analysis, trend identification - Volume pattern analysis, unusual trading activity - Sentiment scoring (Farcaster, influencer activity)

Example Queries (Natural Language → SQL)

Agent asks in natural language; MCP tools discover schema and execute SQL. Example prompts:

Workflow

1. Discover available data: Call list_datasets to see datasets and tables 2. Understand schema: Call get_schema for the dataset you need 3. Write and run: Use query with valid SQL (SELECT, bare table names, LIMIT where appropriate) 4. Interpret results: Use returned data for analysis, proposals, or decisions

Runtime Enforcement (Mandatory)

Before answering with Pond3r-backed data:

1. Prefer scripts when MCP tools are unavailable: - If list_datasets, get_schema, query are not exposed by the runtime, use the CLI scripts instead. - Run node /opt/pond3r-skill-scripts/list-datasets.mjs (or workspace path) with POND3R_API_KEY in env. 2. If neither MCP nor scripts work: - Stop and return: Pond3r unavailable: MCP tools missing and scripts failed (check POND3R_API_KEY in env). 3. Do not assume fallback permission: - Do not switch to web_search, web_fetch, or other sources unless the user explicitly approves fallback. 4. Return execution evidence: - Include the exact commands run and summarize returned dataset/query output. - If a call fails, include the exact error message and next remediation step.

Troubleshooting

| Symptom | Fix | |---------|-----| | "Pond3r MCP not configured" | Add MCP server in your runtime (Cursor/Claude) with URL + Bearer header. Restart. | | Tools still missing after config | Restart the app (Cursor/Claude). MCP loads at startup. | | Agent runs in OpenClaw/Telegram | Use the CLI scripts with POND3R_API_KEY in .env. See CLI Scripts. | | Auth/401 errors | Check API key is valid, not expired. Rotate if it was ever exposed. |

Failure Handling

Report API (Alternative to MCP)

For scheduled reports and structured JSON delivery, use the REST API:

Report format includes executiveSummary, analysis, opportunities. For full API details and response structure, see reference.md.

Security

| Symptom | Fix | |---------|-----| | "Pond3r MCP not configured" | Add MCP server in your runtime (Cursor/Claude) with URL + Bearer header. Restart. | | Tools still missing after config | Restart the app (Cursor/Claude). MCP loads at startup. | | Agent runs in OpenClaw/Telegram | Use the CLI scripts with POND3R_API_KEY in .env. See CLI Scripts. | | Auth/401 errors | Check API key is valid, not expired. Rotate if it was ever exposed. |