Skip to main content

Connection Reference

Every MCP client needs two values to connect to Bloque:

FieldValue
URLhttps://mcp.bloque.run/mcp
AuthorizationBearer <your API key>

Claude (claude.ai, Desktop, and mobile)​

Claude's custom connectors support API-key authentication via a request header, so claude.ai and Claude Desktop can both connect to Bloque without OAuth. See Claude Desktop for the setup steps — the same steps work on claude.ai in a browser.

Claude's mobile apps can use a connector once it's added on web or Desktop, but can't add a new custom connector directly.

Environment Variables​

Some MCP clients read environment variables from the process that starts the client. They usually do not load .env files automatically, so make sure BLOQUE_API_KEY is available before launching the client.

Option 1: Export the key in your shell​

Add this to your shell profile, such as ~/.zshrc or ~/.bashrc:

export BLOQUE_API_KEY="your Bloque API key"

Open a new terminal, then start your MCP client from that terminal.

Option 2: Use a small .env wrapper​

If you keep the key in a project .env file:

BLOQUE_API_KEY="your Bloque API key"

Create a wrapper script such as bin/client-with-env:

#!/usr/bin/env bash
set -euo pipefail

set -a
source .env
set +a

exec "$@"

This works best when .env contains simple shell-compatible KEY=value lines.

Option 3: Read the key from an OS secret store​

Store the key in your operating system's managed keychain, then read it at launch time. For example, with cross-keychain:

#!/usr/bin/env bash
set -euo pipefail

export BLOQUE_API_KEY="$(cross-keychain get bloque BLOQUE_API_KEY)"

Visual Studio Code​

VS Code supports native remote MCP servers over HTTP. No bridge needed.

Edit your global or workspace mcp.json (run MCP: Add Server from the Command Palette, or edit the file directly):

{
  "servers": {
    "bloque": {
      "type": "http",
      "url": "https://mcp.bloque.run/mcp",
      "headers": {
        "Authorization": "Bearer ${input:bloque-api-key}"
      }
    }
  },
  "inputs": [
    {
      "type": "promptString",
      "id": "bloque-api-key",
      "description": "Bloque API key"
    }
  ]
}

Cursor​

  1. Open Cursor → Integrations → MCP.
  2. Choose Edit mcp.json.
  3. Add or merge the bloque entry:
{
  "mcpServers": {
    "bloque": {
      "type": "http",
      "url": "https://mcp.bloque.run/mcp",
      "headers": {
        "Authorization": "Bearer <your API key>"
      }
    }
  }
}

LM Studio​

  1. Open LM Studio → Integrations → MCP and select Edit mcp.json.
  2. Add the bloque entry inside the mcpServers object:
{
  "mcpServers": {
    "bloque": {
      "type": "http",
      "url": "https://mcp.bloque.run/mcp",
      "headers": {
        "Authorization": "Bearer <your API key>"
      }
    }
  }
}

Saving the file automatically reloads MCP servers in LM Studio.

Dify​

Dify has built-in MCP support from v1.6.0. Only HTTP transport servers are supported — Bloque's endpoint qualifies.

  1. Open Tools → MCP and click Add MCP Server (HTTP).
  2. Server URL: https://mcp.bloque.run/mcp
  3. Set a name (e.g. Bloque) and a server identifier (e.g. bloque). Apps reference the server by this identifier, so avoid changing it later.
  4. Under custom headers, add — Name: Authorization, Value: Bearer <your API key>.
  5. Save. Dify connects and imports Bloque's tools, which you can then add to any app or workflow.

If a server behind Bloque streams long-running results, raise the request and SSE read timeouts in the server's settings.

n8n​

Use the MCP Client Tool node (for AI Agent workflows) or the MCP Client node.

  1. Add the node and set the endpoint: https://mcp.bloque.run/mcp
  2. Server Transport: HTTP Streamable (SSE is deprecated).
  3. Authentication: Bearer — create a credential containing your Bloque API key.
  4. Choose which tools to expose to the agent (all, or a selected subset).

Run the workflow; the agent can now call any tool from the servers in your Bloque Hub.