MCP Bridge / Cross-Project Tool Mesh

Massu AI's MCP Bridge lets you call tools on external MCP servers directly from Claude Code. Any project can define servers in a .mcp.json file, and Massu will automatically discover and connect to them — creating a tool mesh across your projects.

Why This Matters

AI coding assistants are powerful within a single project, but real-world development often involves multiple projects, services, and toolchains. MCP Bridge breaks down these silos:

• Cross-project tool access — call tools from your Python backend while working in your TypeScript frontend
• Custom tool integration — expose domain-specific tools (data pipelines, deployment scripts, internal APIs) to Claude Code
• Unified workflow — no context switching between terminals or projects
• Security by default — environment variable filtering, deny patterns, and allow-lists protect sensitive data

Configuration

Define your MCP servers in a .mcp.json file at your project root:

{
  "mcpServers": {
    "my-python-tools": {
      "command": "python",
      "args": ["-m", "my_mcp_server"],
      "cwd": "/path/to/python-project",
      "env": {
        "DATABASE_URL": "postgresql://localhost/mydb"
      }
    },
    "data-pipeline": {
      "command": "node",
      "args": ["./tools/mcp-server.js"],
      "cwd": "/path/to/data-project"
    }
  }
}

Each server entry specifies:

• command — the executable to launch the MCP server
• args — command-line arguments
• cwd — working directory for the server process
• env — environment variables passed to the server (filtered by security rules)

Security

MCP Bridge includes multiple security layers:

Environment Variable Filtering

Sensitive environment variables are automatically filtered before being passed to external servers. Variables matching common secret patterns (_SECRET, _TOKEN, _KEY, _PASSWORD) are stripped unless explicitly allow-listed.

Deny Patterns

Configure deny patterns to block specific tool calls or server connections:

# In massu.config.yaml
mcp_bridge:
  deny_patterns:
    - "dangerous_tool_*"
    - "admin_*"

Allow-Lists

Restrict which servers and tools are accessible:

# In massu.config.yaml
mcp_bridge:
  allow_servers:
    - "my-python-tools"
    - "data-pipeline"
  allow_tools:
    - "query_database"
    - "run_migration"

Tools

massu_mcp_servers

What it does: Lists all MCP servers configured in your project's .mcp.json file. Shows server names, connection types (stdio/SSE), commands, and environment configuration.

Usage:

massu_mcp_servers

Example output:

Configured MCP Servers:
  my-python-tools  stdio  python -m my_mcp_server  (3 env vars)
  data-pipeline    stdio  node ./tools/mcp-server.js  (0 env vars)

Tier: Pro

massu_mcp_tools

What it does: Discovers available tools on any connected MCP server. Shows tool names, descriptions, and input schemas.

Usage:

massu_mcp_tools { "server": "my-python-tools" }

Example output:

Tools on my-python-tools:
  query_database    Query the project database with parameterized SQL
  run_migration     Apply pending Alembic migrations
  seed_data         Insert test data for development

Tier: Pro

massu_mcp_call

What it does: Invokes a tool on an external MCP server with full argument forwarding. Supports stdio and SSE transports with automatic connection management.

Usage:

massu_mcp_call {
  "server": "my-python-tools",
  "tool": "query_database",
  "arguments": {
    "query": "SELECT COUNT(*) FROM users",
    "params": []
  }
}

The tool connects to the specified server, calls the named tool with the provided arguments, and returns the result directly to your Claude Code session.

Tier: Pro

massu_mcp_status

What it does: Checks connection health and latency of all configured MCP servers. Shows uptime, response times, and error counts.

Usage:

massu_mcp_status

Example output:

MCP Server Status:
  my-python-tools  ✓ healthy  latency: 12ms   errors: 0
  data-pipeline    ✓ healthy  latency: 8ms    errors: 0

Tier: Pro

Common Use Cases

Full-Stack Development

Working on a TypeScript frontend but need to check your Python backend's database schema? Use MCP Bridge to call your Python project's tools without leaving your current session:

massu_mcp_call { "server": "backend", "tool": "list_models" }

Shared Internal Tools

Expose your team's internal tools (deployment scripts, data pipelines, monitoring) as MCP servers and access them from any project:

massu_mcp_call { "server": "deploy-tools", "tool": "check_staging" }

Multi-Service Architecture

In a microservices setup, connect to each service's MCP server to understand the full system without cloning every repository:

massu_mcp_tools { "server": "auth-service" }
massu_mcp_tools { "server": "billing-service" }