There is no API product to sign up for here: the server drives the same endpoints the website uses, with your `sessionid` standing in for a key — so anything the model does counts against your own account, and the token has the reach of a logged-in session. Blending mode is the part worth knowing about: supply a reference picture and it uploads and switches modes automatically, which is one call instead of a manual pipeline.
A TypeScript MCP server that calls Jimeng AI's image and video generation directly, authenticating with the session token from your own Jimeng account rather than through a third-party API.
- `generateImage` — text to image, returning a list of image URLs
- Reference-image generation: pass `filePath` with a local path or an image URL and it switches to blending mode, uploading the picture for you
- `generateVideo` — text to video, with reference images for the first and last frames and a resolution setting
- Control over `width`, `height`, `sample_strength` and `negative_prompt` on image calls
- A named model list covering the Seedream image models and Seedance video models, or pass a raw model key straight through
- Multiple tokens separated by commas, rotated in order, for load-balancing across accounts
A Jimeng account. `JIMENG_API_TOKEN` is the `sessionid` cookie value taken from the Jimeng site in your browser's developer tools — treat it like a password. Install through Smithery, or run the built server with Node and set the token in your client config. Reference images passed by URL must be publicly reachable.
One command plus a key — npx -y @smithery/cli install @c-rick/jimeng-mcp --client claude, then supply credentials