Best for
- Developers needing quick visual assets for their projects
- Users who want to generate images directly from Claude Desktop or Cursor
- Scenarios requiring structured output (e.g., JSON) for programmatic use
R
Replicate Flux MCP
Community
MCP server for high-quality image and SVG generation for developers
GitHub source repository ↗
56FMRS · C
This server is feature-rich, supporting multiple image generation modes, but depends on the Replicate API which requires a paid token. Setup is straightforward and integrates well with major clients, but lacks automated tests.
Replicate Flux MCP is an advanced Model Context Protocol (MCP) server that empowers AI assistants to generate high-quality images and vector graphics. By default it uses black-forest-labs/flux-schnell for raster images and recraft-ai/recraft-v3-svg for SVG output. You can override the curated image/SVG models through environment variables or per-tool model_id overrides (from built-in allowlists). It supports batch image generation, variant generation, arbitrary Replicate model invocation, structured output, progress notifications, and works with Claude Desktop, Cursor, Cline, Zed, and other MCP-compatible clients.
Tools
generate_image
Generates an image based on a text prompt
generate_multiple_images
Generates multiple images from an array of prompts
generate_image_variants
Generates multiple variants of an image from a single prompt
generate_svg
Generates SVG/vector output based on a text prompt
prediction_list
Retrieves a list of recent predictions from Replicate
get_prediction
Gets detailed information about a specific prediction
run_model
Runs a whitelisted Replicate model with a raw input payload
run_replicate_model
Runs any model hosted on Replicate by its owner/name reference
get_model_schema
Fetches the OpenAPI input schema and description for a Replicate model
Setup
- Obtain a Replicate API token by signing up at replicate.com. 2. Choose your integration method: for Claude Desktop, add the configuration to your mcp.json file; for Cursor, create/edit .cursor/mcp.json; for Codex, add to ~/.codex/config.toml. 3. Replace 'YOUR_TOKEN' with your actual API token. 4. Restart your MCP client to apply changes.
claude_desktop_config.json
{
"mcpServers": {
"replicate-flux-mcp": {
"command": "npx",
"args": [
"-y",
"replicate-flux-mcp"
],
"env": {
"REPLICATE_API_TOKEN": "YOUR TOKEN"
}
}
}
}
Fit and risk
Not for
- Fully offline image generation (relies on Replicate API)
- Use cases with strict privacy requirements where prompts cannot be sent to external services
- Users expecting unlimited free generation (Replicate API is usage-based and paid)
Required permissions
- Access to Replicate API (requires REPLICATE_API_TOKEN)
- Making network requests to Replicate to generate and retrieve results
- Reading and writing local configuration files and environment variables
Risks and side effects
- API key leakage could lead to unauthorized usage and charges
- Generated images may contain inappropriate content (safety filter exists but can be bypassed)
- Dependence on external service: network latency or downtime possible
- Running arbitrary models via run_replicate_model may incur unexpected costs and outputs
Troubleshooting
- Authentication error: verify REPLICATE_API_TOKEN is correctly set and valid
- Safety filter triggered: modify prompt to avoid problematic content
- Timeout error: increase pollingAttempts or pollingInterval in configuration
- Model override not working: ensure model ID is in the built-in allowlist
Use cases
Generating custom images and SVG icons for coding projects
Batch-generating multiple concepts to explore design options
Creating multiple style variants of the same prompt (e.g., watercolor, oil painting)
Invoking arbitrary Replicate models for video generation, image upscaling, etc.
Supported clients
Claude DesktopFull support
CursorFull support
ClineFull support
ZedFull support
CodexFull support