PixelDojo MCP¶
Media generation behind one endpoint. PixelDojo gives an MCP client access to 140+ generative models for images, video and audio, plus the studio workflows built on top of them. An agent can generate a still, edit or upscale it, cut a product video ad, run a multi-step campaign, keep a character consistent across shots, or produce a faceless explainer episode and a music video, with the model chosen per prompt and a download link returned.
Server type: Remote (Streamable HTTP) and local (stdio)
Auth: OAuth 2.0 browser sign-in, or a Bearer API key (keys start with pd_)
Endpoint: https://pixeldojo.ai/mcp
Local: npx -y @pixeldojo/mcp (v0.16.3)
Tools: 19 in the server, 18 over the hosted endpoint
Pricing: Lite 10 USD per month (160 credits), Pro 25 USD (500), Pro Max 50 USD (1,100)
Category: Creative
Built by: BLOVE INC, d/b/a Pixel Dojo (Brian Lovett)
What it exposes¶
Nineteen tools ship in the server. The planning tools are free: pixeldojo_from_url takes a public media URL as input, and the estimate step inside pixeldojo_film and pixeldojo_episode costs nothing.
Generation and editing cover the single-shot work: pixeldojo_generate, pixeldojo_edit, pixeldojo_character, pixeldojo_upscale, pixeldojo_audio and pixeldojo_video.
The studio tools wrap multi-step pipelines: pixeldojo_storyboard and pixeldojo_ad for product video ads, pixeldojo_campaign with pixeldojo_campaign_status for a full campaign, and pixeldojo_film for short films.
Workflow and library tools make the output reusable: pixeldojo_save_workflow, pixeldojo_run_workflow, pixeldojo_list_workflows, pixeldojo_status for long-running jobs and pixeldojo_library for finished assets.
One tool is stdio-only: pixeldojo_upload reads from the local filesystem, so the hosted endpoint does not offer it. That leaves 18 tools over https://pixeldojo.ai/mcp, a fact worth knowing if a directory listing reports fewer. Third-party directories generally cannot enumerate the tool list at all, because the endpoint refuses unauthenticated calls.
Connecting¶
For Claude Code, add the hosted transport:
claude mcp add --transport http pixeldojo https://pixeldojo.ai/mcp
Then sign in through the browser when the client prompts, or paste an API key created at pixeldojo.ai in the API platform section. Any other client that speaks Streamable HTTP takes the same URL; the endpoint advertises OAuth protected-resource metadata at https://pixeldojo.ai/.well-known/oauth-protected-resource/mcp, so clients with automatic discovery handle the sign-in themselves.
For a local run, npx -y @pixeldojo/mcp starts the stdio server on Node 18 or newer, reading PIXELDOJO_API_KEY from the environment. Running locally is the only way to get pixeldojo_upload and the only way to feed the server a file that is not already reachable on the public web.
Notes¶
Asset URLs expire one hour after creation, so anything worth keeping should be downloaded immediately or left in the library, where finished films persist. Generation calls wait 30 seconds before returning a jobId; a longer render is picked up with pixeldojo_status rather than by holding the connection open.
Pricing is credit-based and every call spends credits, with failed jobs refunded. Model rosters at this layer move quickly, so treat the specific model list as a snapshot rather than a fixed catalogue. The server reports its own origin on each request through a source header, which is useful when auditing where generation spend came from.
Integration with CorpusIQ¶
CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.
This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.