Connect the MCP to Claude & other clients

Generate your credentials

MCP credentials are generated from the BigIdeasDB dashboard and require a Pro membership. Each credential is a pair: a client ID that starts with biddb_ and a client secret that starts with biddb_sec_. The secret is shown only once when you generate it, so copy both somewhere safe before you leave the page. The server URL is the same everywhere; only the credential pair is unique to you.

https://bigideasdb.com/api/mcp/bigideasdb-mcp

The universal config

Most clients accept a hosted MCP server over HTTP with a JSON config like this. Replace your_client_id and your_client_secret with the pair you generated in the dashboard. Both headers are required; the server rejects a request that is missing either one. Wherever a client uses a JSON MCP config file, this is the shape you paste in.

{
  "mcpServers": {
    "bigideasdb-mcp": {
      "type": "http",
      "url": "https://bigideasdb.com/api/mcp/bigideasdb-mcp",
      "headers": {
        "X-Client-Id": "your_client_id",
        "X-Client-Secret": "your_client_secret"
      }
    }
  }
}

Claude Code and Claude Desktop

In Claude Code you can add the hosted server from the command line, giving it a name, the URL, and your two credential headers. After adding it, reload so the tools register.

claude mcp add bigideasdb-mcp --transport http https://bigideasdb.com/api/mcp/bigideasdb-mcp \
  --header "X-Client-Id: your_client_id" \
  --header "X-Client-Secret: your_client_secret"

Claude Desktop reads its servers from the app's configuration file. Add the bigideasdb entry to the mcpServers object and restart the app so it picks up the change.

{
  "mcpServers": {
    "bigideasdb-mcp": {
      "type": "http",
      "url": "https://bigideasdb.com/api/mcp/bigideasdb-mcp",
      "headers": {
        "X-Client-Id": "your_client_id",
        "X-Client-Secret": "your_client_secret"
      }
    }
  }
}

Cursor, VS Code, and Windsurf

Editor-based clients expose MCP through a settings pane or a JSON file in the project or user directory. Cursor and Windsurf connect through the mcp-remote bridge, which forwards your credential headers to the hosted server. Add the entry in the MCP settings and reload the editor.

{
  "mcpServers": {
    "bigideasdb-mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://bigideasdb.com/api/mcp/bigideasdb-mcp",
        "--header",
        "X-Client-Id: your_client_id",
        "--header",
        "X-Client-Secret: your_client_secret"
      ]
    }
  }
}

VS Code with GitHub Copilot registers hosted MCP servers in its own MCP configuration. The server object uses the same URL and headers; after adding it, reload the window so the bigideasdb tools become available to the chat.

{
  "servers": {
    "bigideasdb-mcp": {
      "type": "http",
      "url": "https://bigideasdb.com/api/mcp/bigideasdb-mcp",
      "headers": {
        "X-Client-Id": "your_client_id",
        "X-Client-Secret": "your_client_secret"
      }
    }
  }
}

ChatGPT and other clients

ChatGPT connects to hosted MCP servers as a connector. Point it at the server URL and supply your credential as two custom headers, X-Client-Id and X-Client-Secret. The same pattern applies to any other client that speaks hosted MCP over HTTP, including Codex, JetBrains AI, and Replit.

• Server name - bigideasdb-mcp (any label you like).
• Server URL - https://bigideasdb.com/api/mcp/bigideasdb-mcp
• Authentication - two headers, X-Client-Id and X-Client-Secret, set to your credential pair.

name: bigideasdb-mcp
url:  https://bigideasdb.com/api/mcp/bigideasdb-mcp
headers:
  X-Client-Id: your_client_id
  X-Client-Secret: your_client_secret

For the full list of confirmed clients and the account-wide rate limits, see Rate limits and supported clients.

Verify the connection

After adding the server, restart or reload your client and confirm the bigideasdb tools appear. Then try a prompt like "search Reddit for complaints about invoicing software." If tools do not appear, double-check both credential headers and the server URL, or see Help.

Good first prompts that exercise different sources so you know the connection is healthy:

"Search Reddit for people frustrated with scheduling apps."
"Show me Capterra feature gaps for project management software."
"How many companies are live on Stripe in the invoicing category?"

If a prompt returns a sourced answer, the connection is working and the model is calling the tools. Because a single prompt can trigger several tool calls, keep an eye on volume during heavy sessions - the limits are shared across every client you connect.

Troubleshooting

• Tools missing - confirm you fully reloaded or restarted the client after editing the config.
• Auth errors - a 401 saying "Missing X-Client-Id or X-Client-Secret header" means the headers did not arrive; check both are set and spelled exactly. An Authorization bearer header does not work here.
• Wrong URL - the server URL must be exactly https://bigideasdb.com/api/mcp/bigideasdb-mcp with no trailing path.
• Still failing - generate a fresh credential in the dashboard and paste the new pair in.

Managing your credentials across clients

One credential pair works everywhere, so you do not need a separate credential per client. If you research from Claude Code during the day and switch to Cursor or ChatGPT later, paste the same client ID and secret into each and they all authenticate against the same account. That keeps setup simple, but it also means every connected client draws down one shared pool of requests.

You can hold up to three active credentials at once, and generating a new one does not disable the old ones. To retire a credential, revoke it in the dashboard, which stops it working immediately. When you rotate, update both headers in each client you have configured, then reload so the new credential takes effect. Storing the secret in an environment variable rather than a committed config file makes rotation painless and keeps it out of version control.

The account-wide limits that every client shares are covered in full on Rate limits and supported clients.