Creating an Amazon Bedrock Nova Canvas MCP server and integrating it with Claude Desktop
In this article, we explain how to create a local MCP server that uses AWS resources to invoke the Amazon Nova Canvas model through Claude Desktop and generate images from natural language. This guide includes practical steps, basic commands, and reference links to obtain the main files.
Step 1: Prepare the local Python project
Run the following commands in your terminal to create the project and the virtual environment uv init mcp-nova-canvas cd mcp-nova-canvas uv venv source .venv/bin/activate uv add mcp[cli]
Step 2: Configure dependencies
In the pyproject.toml file, define the project metadata and add key dependencies such as boto3, httpx, mcp[cli], pillow, uuid, loguru, and Python version 3.11 or higher. You can adapt the versions according to your security policies and AWS compatibility.
Step 3: Add the main files
Copy the content of main.py from the reference repository on GitHub and create the utils.py and constants.py files according to the examples. Main link for reference https://github.com/Gediontkd/mcp-nova-canvas/blob/main/main.py Utils link https://github.com/Gediontkd/mcp-nova-canvas/blob/main/utils.py Constants link https://github.com/Gediontkd/mcp-nova-canvas/blob/main/constants.py
Step 4: Configure Claude Desktop to detect the MCP
Edit the claude_desktop_config.json file in the Claude Desktop configuration to add an MCP server that invokes the uv command. The entry should specify the command uv and arguments such as --directory absolute_path_to_project run main.py. Make sure to replace absolute_path_to_project with the full path to the mcp-nova-canvas directory on your machine. If you cannot find the file, open Claude Desktop and go to Settings > Developer > Edit Config to locate it.
Step 5: Restart Claude Desktop and generate images
Close Claude Desktop with File > Exit and reopen it. In the Claude Desktop interface, use the text command generate_image followed by your description prompt to request image generation. For example, generate_image a classic Italian restaurant scene at sunset with warm lights. The MCP server will send the request to Amazon Nova Canvas and download the result to the project output folder.
Step 6: Verify the result
Open the output folder in the project main directory and check the generated image files. If you need to adjust generation parameters or image size, edit the logic in main.py or utils.py and restart the server from Claude Desktop.
Tips and best practices
Configure AWS credentials with the minimum permissions required for Bedrock and storage if you use S3. Apply security practices such as credential rotation and access control. Test short prompts and gradually increase complexity to optimize cost and latency.
Summary
In this module, you learned how to build an MCP server in a local environment and integrate it with Claude Desktop to invoke Amazon Nova Canvas and generate images using natural language. The steps cover creating the Python project, configuring dependencies, adding example files, configuring Claude Desktop, and verifying the results in the output folder.
About Q2BSTUDIO
Q2BSTUDIO is a custom software and application development company specialized in enterprise solutions that integrate artificial intelligence, cybersecurity, and AWS and Azure cloud services. We offer custom software, custom applications, AI agents, and business intelligence solutions and Power BI implementation to transform data into decisions. Our team of artificial intelligence specialists develops AI for companies that require automation, analytics, and secure conversational AI agents. We also manage AWS and Azure cloud services and cybersecurity solutions tailored to each client.
Keywords for positioning
custom applications, custom software, artificial intelligence, cybersecurity, AWS cloud services, Azure cloud services, business intelligence services, AI for companies, AI agents, Power BI
Contact and resources
LinkedIn https://www.linkedin.com/in/gedion-daniel-760ba6280 GitHub https://github.com/Gediontkd Website https://www.gediondaniel.dev For professional custom software development services and AI and cloud solution integration, contact Q2BSTUDIO and request a personalized technical consultation.
