Bedrock Image MCP – AI Image Generation and Editing
The Bedrock Image MCP server connects AI agents to supported Amazon Bedrock image-generation and image-editing models. Use it for creating visual assets, transforming reference images, applying controlled edits, generating masks, and exporting production-ready image variants from MCP-compatible clients.
Overview
The Bedrock Image MCP server is a community-maintained image-generation and editing integration for Amazon Bedrock. It began as a fork of the AWS Labs Nova Canvas MCP server and extends that foundation with additional Amazon Nova Canvas, Stable Diffusion 3.5 Large, and Stability AI image-service tools. It gives MCP-compatible AI agents structured access to visual-generation and editing operations instead of requiring users to call Bedrock model APIs manually.
What the MCP server enables
Depending on model availability, AWS permissions, and regional support, an AI agent can:
- Generate images from text prompts with Amazon Nova Canvas.
- Guide Nova Canvas output with selected color palettes.
- Create images with Stable Diffusion 3.5 Large using supported aspect ratios and seeds.
- Transform existing images with image-to-image generation.
- Upscale images with creative, conservative, or fast workflows.
- Inpaint masked regions and outpaint beyond the original image boundaries.
- Search for and replace or recolor selected objects.
- Remove objects or backgrounds from images.
- Convert sketches or structural guides into finished imagery.
- Apply reference styles and perform style transfer.
- Create rectangular, elliptical, or full-image masks for edit operations.
- Save outputs in PNG, JPEG, or WebP formats to a chosen workspace directory.
When to use it
Use this MCP server when an AI-assisted design workflow needs repeatable image production through Amazon Bedrock. Typical uses include campaign concepts, product mockups, UI illustrations, social assets, background cleanup, object replacement, style exploration, resolution enhancement, and generating multiple controlled visual variants from prompts or reference images.
Connection and authentication
The documented local connection uses stdio and the Python package runner uvx:
uvx bedrock-image-mcp-server@latest
The server authenticates through the standard AWS credential chain. Its documented configuration supports AWS_PROFILE and AWS_REGION; the selected IAM user or role must have permission to invoke the required Amazon Bedrock foundation models. Credentials remain outside this YAML and should be stored in AWS configuration, environment variables, or another approved credential provider.
Key considerations
This is a community-maintained server rather than an AWS-operated hosted MCP endpoint. Model access, regions, pricing, quotas, content filtering, and supported features can change independently. Grant only the required bedrock:InvokeModel permissions, keep output directories controlled, and review generated or edited images before publication. Input dimensions, masks, prompts, and aspect ratios must follow the requirements of the selected model. The older AWS Labs Nova Canvas MCP server was deprecated in March 2026 in favor of this broader community-maintained replacement.
Supported Transports
stdio
Command: uvx
Args:
bedrock-image-mcp-server@latest
Frequently Asked Questions
- When should an AI agent use the Bedrock Image MCP server?
- Use it when a design workflow needs image generation or editing through Amazon Bedrock, including concept art, campaign assets, product mockups, image transformation, upscaling, background removal, inpainting, outpainting, or controlled style variations.
- What does the Bedrock Image MCP server add to an AI agent's capabilities?
- It gives the agent structured tools for supported Amazon Nova Canvas, Stable Diffusion 3.5 Large, and Stability AI image operations, allowing the agent to generate and edit image files rather than only describe visual changes.
- What can an AI agent access or manage through this server?
- The agent can use documented tools for text-to-image generation, color-guided generation, image-to-image transformation, upscaling, inpainting, outpainting, object replacement, recoloring, object and background removal, sketch or structure control, style guidance, style transfer, mask creation, and local output management.
- How is authentication configured for the Bedrock Image MCP server?
- The server uses the standard AWS credential chain and can select a named profile with AWS_PROFILE and a target region with AWS_REGION. The IAM identity must have bedrock:InvokeModel access to each required model, and credentials should remain in AWS configuration or a secure credential provider rather than this YAML.
- Which transport should be used for the Bedrock Image MCP server?
- Use the documented local stdio transport launched with uvx and the bedrock-image-mcp-server package. The reviewed project documentation does not provide a provider-hosted Streamable HTTP or SSE endpoint, so only stdio is included.