Strands

Experimental: Bidirectional MCP integration for Strands Agents.

python

7

GitHub Stars

python

Language

6 months ago

First Indexed

2 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "cagataycali-strands-mcp-server": {
      "command": "uvx",
      "args": [
        "strands-mcp-server",
        "--cwd",
        "/path/to/project"
      ]
    }
  }
}

You can expose your Strands Agent as an MCP server to enable bidirectional communication with MCP clients. This lets other tools and agents call your exposed tools over HTTP or stdio, making your Strands setup interoperable with Claude Desktop, Kiro, and other MCP-enabled clients.

How to use

To use the MCP server, start the local MCP service within your Strands environment and connect a client to it. You can expose specific tools or expose all available tools from your agent, and you can choose between HTTP or stdio transports depending on your integration needs. Once the server is running, a client can request to call a tool, submit inputs, and receive results.

A typical workflow is to run the MCP server locally and then configure a client to connect either over HTTP or via stdio. For Claude Desktop or other MCP-enabled clients, you provide a configuration that points to your local server so the client can invoke your exposed tools as if they were local functions.

How to install

Prerequisites: ensure Python and pip are installed on your system.

Install the MCP server package with Python’s package manager.

pip install strands-mcp-server

Configuring a client to connect

To connect a client to a remote MCP server, configure the client with the server URL and specify how to transport messages (HTTP or stdio). The client can then request to call tools exposed by the server.

# Example client usage (abstracted; adapt to your client framework)
# Connect to a server over HTTP
# transport = 'http'
# server_url = 'http://localhost:8000/mcp'
# Then call a tool by name with arguments

Additional sections

CLI usage mirrors the standard command line interface for running the MCP server in local mode or as a proxy to another MCP endpoint.

If you are integrating with Claude Desktop or similar clients, you can provide a JSON configuration that points to your local MCP server and specifies how it should start. The exact fields depend on your environment but typically include a command to start the MCP server and any necessary working directory paths.

Troubleshooting

If you encounter connection issues, verify that the MCP server is running and that the client is pointing to the correct URL or stdio endpoint. Check that the port is not in use by another process and review relevant logs for errors related to authentication, transport, or tool invocation.

Available tools

mcp_server

Exposes the server side that makes Strands tools available to MCP clients as an MCP server.

mcp_client

Client-side tool to connect to MCP servers and invoke remote tools.

cli

Command line interface for running the MCP server locally or proxying to remote MCP endpoints.