Docs / Quickstart

Quickstart

Connect any MCP-capable agent to CoinFactory in under a minute. No account, no API key.

MCP SERVER URL — STREAMABLE HTTP, STATELESS, NO AUTH

→https://coinfactory.ai/mcp

Claude Code

terminal

claude mcp add --transport http coinfactory https://coinfactory.ai/mcp

# then, in any session:
> what's trending on base right now, and how risky is the top one?

Claude Desktop

Settings → Connectors → Add custom connector, paste the URL above. Or edit claude_desktop_config.json:

claude_desktop_config.json

{
  "mcpServers": {
    "coinfactory": { "url": "https://coinfactory.ai/mcp" }
  }
}

Cursor

.cursor/mcp.json

{
  "mcpServers": {
    "coinfactory": { "url": "https://coinfactory.ai/mcp" }
  }
}

OpenClaw and other agents

Any agent that speaks MCP over HTTP can use the URL directly. Agents that can only read web pages should start with the skill file — it contains the same instructions in Markdown, including the REST fallbacks:

prompt

Read https://coinfactory.ai/skill.md and follow the instructions.

Verify with MCP Inspector

terminal

npx @modelcontextprotocol/inspector --transport http --server-url https://coinfactory.ai/mcp

# or raw JSON-RPC:
curl -s https://coinfactory.ai/mcp \
  -H 'content-type: application/json' -H 'accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"get_risk_signals","arguments":{"token":"pepe"}}}'

Recommended agent workflow: search_tokens to resolve → get_token for numbers → get_risk_signals before forming any opinion. Quote the signal explanations, not just the score. Never describe a token as "safe".