What PixelDojo MCP server does
PixelDojo MCP server exposes PixelDojo’s generative media platform to AI agents through MCP. It covers image, video, audio, and short-film production using a catalog of more than 140 models, including models identified in the source material as Flux, Seedream, Seedance, Veo, Kling, WAN, Nano Banana, GPT Image, and Seed Audio.
The server supports both general generation and focused production workflows. Agents can create images or videos from prompts, edit one or more existing images, maintain a consistent character across shots, produce individual storyboard shots, or assemble multi-shot films and music videos. Marketing and Faceless Studio tools can turn product information or a topic into video content. Product campaigns can produce a hero image, lifestyle images, and an optional video from one product URL.
Additional tools cover product-page extraction, image and video upscaling, combined dialogue/effects/music audio, saved-media search, reusable workflows, and long-running job status checks.
How it works
The hosted service is available at https://pixeldojo.ai/mcp over streamable HTTP. Users authenticate through browser approval, and the hosted endpoint can also accept a PixelDojo bearer API key. The local distribution runs with npx -y @pixeldojo/mcp over stdio and requires an API key in the client environment.
A prompt can be sent to the general generation tool, or an agent can choose a specialized tool such as pixeldojo_ad, pixeldojo_film, or pixeldojo_episode. PixelDojo selects a model from its catalog unless a model identifier is supplied. Requests that finish quickly return their result directly. Longer renders return a jobId, which the agent can pass to pixeldojo_status.
Generated asset URLs remain available for one hour, so assets that need to be retained should be downloaded promptly. Completed films remain in the PixelDojo library. Rendering consumes credits from the associated PixelDojo account, and failed jobs are refunded.
Setup and configuration
For the hosted connection, configure the MCP URL in the client and complete browser sign-in. The README provides connection examples for Claude Code, Claude Desktop or claude.ai, Cursor, VS Code, and Windsurf. For a local connection, create an API key through PixelDojo, run the package with npx, and set PIXELDOJO_API_KEY in the MCP server configuration.
The hosted server exposes 18 tools. It does not expose pixeldojo_upload because the hosted process cannot read local files. The local upload tool converts a local file into a temporary URL that lasts 24 hours. API keys are redacted from logs. Requests include headers identifying the MCP source and selected skill for traffic attribution.
Tools and capabilities
PixelDojo MCP server includes these tool groups:
pixeldojo_generate, pixeldojo_edit, pixeldojo_character, and pixeldojo_storyboard for image, video, editing, character, and shot creation.
pixeldojo_film, pixeldojo_episode, pixeldojo_ad, and pixeldojo_campaign for films, narrated faceless episodes, product ads, and product campaigns.
pixeldojo_from_url, pixeldojo_upscale, pixeldojo_audio, and pixeldojo_video for product extraction, enhancement, audio, and clip changes.
pixeldojo_library for saved-media and recent-generation searches.
pixeldojo_save_workflow, pixeldojo_run_workflow, and pixeldojo_list_workflows for named multi-step workflows.
pixeldojo_upload for local files in stdio mode and pixeldojo_status for asynchronous jobs.