Kaitoi Studio MCP Server

AI changed creation. Kaitoi changes production.

Kaitoi Studio is a hybrid intelligence platform for creative work — a visual canvas where you design workflows that adapt to your vision, then scale them into real production pipelines.

This MCP server exposes Kaitoi Studio's workflow engine to your AI agent. Build a color-grade pipeline from Claude. Have Cursor wire a Fal video model into an existing graph. Ask your agent to scaffold a custom node, validate it, and drop it onto your canvas. Your agent and you share one workspace — the same canvas you work in visually.

Kaitoi Studio is currently in Private Beta. Join the waitlist →

What You Can Do

Connect your AI agent (Claude Code, Claude Desktop, Cursor, VS Code, Windsurf, or any MCP client) to Kaitoi Studio and drive your creative pipelines from a conversation.

Run workflows

• run_node — Execute a single node and wait for its output
• run_graph — Execute the full graph or a target subset of nodes
• validate_graph — Surface errors and missing connections before you run

Build graphs programmatically

• add_node — Drop any node from the library onto your canvas
• connect_nodes / disconnect_nodes — Wire outputs to inputs
• update_node — Change inputs, parameters, prompts
• delete_node / clear_graph — Prune or reset

Discover nodes

• search_nodes — Semantic + lexical search across builtin, user, and community nodes
• list_packages / get_package / list_package_files / read_package_file — Browse node packages

Inspect state

• get_graph_state — Full structure: nodes, connections, parameters
• get_node_info — Details of a single node
• get_node_output — Fetch outputs from a node that already ran
• get_app_health — System status

Manage projects

• list_projects / load_project / save_project — Persist and recall workflows

Work with files and assets

• browse_library — Search your uploaded files by folder, keyword, or data type
• upload_file — Ingest a file from a URL into your Library
• read_file / write_file / update_file — Edit custom node scripts and other project files

Author custom nodes

• get_node_code — Read the source of any node on your canvas
• validate_node_script — Syntax and schema check against Kaitoi's node spec
• search_user_scripts — Find your custom nodes

Research from inside the agent

• get_fal_docs — Fetch documentation for any fal.ai endpoint (OpenAPI, markdown, and client snippets)
• web_search — Web search that returns URLs, titles, and snippets
• crawl_url — Clean markdown extraction with optional screenshot and link graph

Plus: 10 knowledge resources (node development, runtime API, data types, graph operations, troubleshooting, platform capabilities) and 3 expert prompts (build_computational_workflow, create_custom_node, debug_workflow_issues) your agent can invoke on demand.

Why This Matters

Kaitoi Studio gives your agent something more useful than a single model endpoint: a graph of nodes it can search, compose, and run. Instead of picking one provider and one output type at a time, your agent can work inside a reusable workflow system that already knows how to generate, transform, and combine different kinds of media.

Any provider or model can become part of the workflow. Kaitoi graphs can mix nodes backed by Fal.ai, OpenAI, Anthropic, Google Gemini, OpenRouter, Stability AI, Replicate, Runway, Tripo3D, ElevenLabs, Freepik, Ollama, ComfyUI, and more. Through MCP, your agent can search for the right nodes, wire them together, tune parameters, validate the graph, and run the whole pipeline from one place.

Workflows are reusable, not one-off prompts. Your agent can load existing Kaitoi projects, build on top of established graph p

... View full README on GitHub