Picturesque MCP for any AI

FAQ

How does Picturesque connect to AI agents?

Picturesque exposes an HTTP MCP server. Add the endpoint in Cursor or Claude Code, authenticate with your Picturesque Bearer token, and your agent can call tools like generate_image, list_models, and import_skills in the same session.

Which clients are supported?

Any MCP-compatible client works — Cursor, Claude Code, Claude Desktop, ChatGPT (Developer mode), OpenClaw, Hermes Agent, and custom agents. Paste the server URL and Authorization header into your client config.

What models are available?

list_models returns the live Picturesque catalog — Kling, Veo, Seedance, Nano Banana Pro, GPT Image 2, Suno, Flux, Grok Imagine, and more. Your agent can pick the best model or you can specify one in the prompt.

What can I create?

Images, video, audio, upscales, and brand-aware workflows. list_skills exposes Director slash commands like /cinematic and /ugc. Generate from text, reference images, or both.

Do I need a separate API key?

No separate API key. Sign in at aipicturesque.com, copy your session Bearer token from Director → MCP setup, and pass it in the MCP Authorization header.

How does pricing work?

MCP generations use the same Picturesque credits as the website. Call estimate_generation_cost before submitting a job to preview the charge.

How long does generation take?

Images typically complete in a few seconds. Videos take longer depending on duration and model. All generation runs asynchronously — your agent polls get_task_status and delivers results when ready.

Can I use previous generations as input?

Yes. list_recent_generations lets your agent browse recent outputs and reference past images or video as input for iterative, multi-step production.

Can I sync brand memory and skills?

Yes. upsert_brand_memory writes tone, palette, and guidelines to Director. import_skills uploads custom slash-command workflows you can reuse in chat and external agents.