Skip to main content

Plug the Cryptohopper MCP into n8n, Zapier, or Make

Learn how to connect the Cryptohopper MCP to n8n, Zapier, or Make — with exact node configurations, prompt templates, and fixes for the most common setup issues.

Written by Isaac


Prerequisites

  • A Cryptohopper MCP API key — see how to get a Cryptohopper MCP API key.

  • An account on n8n (self-hosted or cloud), Zapier, or Make.

  • A destination for the output (Telegram, Discord, email, Google Sheets, etc.) — usually already set up if you're on one of these platforms.


Setup — n8n

n8n has the most flexible MCP support of the three platforms.

  1. Add an AI Agent node

    The model can be OpenAI, Anthropic, or any other supported provider.

  2. Add an MCP Client tool under the agent's Tools configuration

    Server URL: https://mcp-data.cryptohopper.com/mcp
    Authentication: Bearer token
    Token: {{ $env.CRYPTOHOPPER_MCP_KEY }} (recommended) or paste directly for local-only environments.

  3. Configure the trigger

    Schedule for daily digests or hourly scans. Webhook for externally-triggered runs (Slack slash command, Discord bot). New-item triggers for RSS, Airtable, or email sources.

  4. Write the prompt in the AI Agent node

    Use a fixed string or build it from trigger data. Example for a daily digest:

Using the Cryptohopper MCP, produce a morning digest for BTC, ETH, SOL, ARB, and OP on Binance. For each token: current price, 24h change, 4h trend, volume state, one-line note.

  1. Add output nodes

    Telegram, Discord, Gmail, Google Sheets — whatever fits your delivery needs.

  2. Test with Execute Workflow

    Verify the MCP tool returns data in the execution log and the delivery node succeeds, then activate for scheduled execution.


Setup — Zapier

  1. Create a new Zap

  2. Add a trigger

    Use Schedule by Zapier or any other trigger source.

  3. Add a Claude or ChatGPT action step

    Both support MCP tool-calling.

  4. Add the MCP server under Tools or MCP Servers

    URL: https://mcp-data.cryptohopper.com/mcp
    Auth header: Authorization: Bearer YOUR_MCP_KEY

  5. Write the prompt in the AI step

    Same shape as the n8n example above.

  6. Add delivery actions and test end-to-end

    Turn the Zap on when working.

Zapier's AI steps meter by task usage. A multi-call MCP workflow may consume several tasks per run — check your plan's limits.


Setup — Make

  1. Create a new scenario

  2. Add a trigger module

    Scheduler, webhook, or a data source.

  3. Add an OpenAI or Anthropic module

    Make labels these "ChatGPT" and "Claude" respectively.

  4. Add the MCP server under Advanced settings → Tools

    Endpoint: https://mcp-data.cryptohopper.com/mcp
    Authorization: Bearer YOUR_MCP_KEY

  5. Write the prompt and add delivery modules

    Telegram, Discord, Email, Airtable — whatever you need.

  6. Schedule the scenario

    Right-click the scenario → Schedule.


Choosing the right platform

Platform

Best for

Key tradeoff

n8n

Complex workflows, self-hosting, full control

More technical setup required

Zapier

Simple workflows, tools already on Zapier

Task-based pricing adds up on frequent runs

Make

Visual clarity, branching logic, mid-complexity workflows

Cloud-only; less flexible than n8n

For most users who also need Google Sheets, Telegram, or Slack integration: n8n if you can self-host, Zapier if you want zero setup friction.


Troubleshooting

The AI step doesn't see the MCP tools

Most commonly a malformed bearer token — confirm the Authorization header is exactly Bearer <key> with a single space, no extra characters. Also confirm your platform's AI step version supports MCP tool-calling.

Every run uses more tasks or credits than expected

A multi-step agent reasoning pass can consume more tasks than you expect. Optimise: simpler prompts, fewer tool calls per run, longer schedule intervals.

Scheduled runs fire but produce empty output

The MCP call likely timed out or returned an error that the agent didn't surface. Add to your prompt: "At the start of your response, say which MCP tools you successfully called; at the end, report any errors you encountered." Makes debugging much faster.

Cryptohopper MCP rate limits hit from the platform

Workflows firing at short intervals can hit the short-interval rate limit. Widen the schedule or split the workload into batches. See rate limits explained.

Output format is inconsistent run to run

Platforms route AI output through their own formatters — Markdown may be stripped, tables mangled. Test end-to-end rendering in the actual delivery channel (not just the platform preview) and adjust your prompt to produce output that renders correctly there.

You want to chain multiple MCPs in one workflow

n8n handles this best — add multiple MCP Client tools to the same agent node. Zapier and Make are more constrained. See combine the Cryptohopper MCP with on-chain data for deeply-stacked workflows.

Your MCP key gets committed to version control via the n8n export

n8n exports include configuration. If you paste keys directly instead of using env references, they land in the export file. Always use {{ $env.VAR }} references and manage keys outside the workflow file.

Did this answer your question?