MCP Canvas:
AI Agents on Your Knowledge Base

Model Context Protocol lets AI agents read and write external tools. Levelry is a visual canvas with native MCP support — your agents can navigate, query, and update your knowledge base directly. No plugins. No workarounds.

Start for free See full features

What is MCP?

The Model Context Protocol is an open standard that lets AI assistants connect to external data sources and tools. Instead of copy-pasting context into a chat window, your agent calls an MCP server directly — reading files, querying databases, updating documents.

Most MCP integrations target text: your agent reads markdown, edits a Notion page, or searches a codebase. Levelry is different. It exposes a visual canvas through MCP — your agent doesn't just read text, it navigates objects and connections in space.

What you can do with a canvas MCP server

Read canvas structure: Agents list objects, their documents, and connections between them.
Query by layer: Separate narrative from mechanics, economy from lore — agents query the layer they need.
Create and update objects: Your agent writes new knowledge directly onto the canvas, not into a text file.
Follow connections: Agents traverse links between objects, understanding relationships that text alone can't express.
Automate workflows: Compile research, generate game design documents, update project status — all through agent-driven MCP calls.

SKILL.md — instant agent integration

We ship a ready-made SKILL.md with the Levelry MCP server. Drop it into any agentic tool — Claude Code, Cursor, Copilot, Hermes, or any agent that reads skill files — and your agent immediately knows how to navigate, read, and update your Levelry canvas.

No custom prompts. No RAG pipeline. The SKILL.md tells your agent exactly what MCP tools are available, when to use them, and how to handle canvas context.

        
          # Quick start

          1. Connect MCP server in your agent config

          2. Add SKILL.md from github.com/kt3v/levelry-docs

          3. Ask your agent to read or update your canvas

          # Example prompts

          "What objects are on my Game Design layer?"

          "Add a new mechanic node connected to Combat System"

          "Summarize all objects on the Lore layer"

View integration docs on GitHub

MCP canvas vs MCP text tools

Most MCP servers expose text — files, documents, search results. Levelry exposes a spatial knowledge graph. Here's the difference:

	Text-based MCP	Levelry MCP canvas
Data model	Flat text / files	Objects with spatial positions
Relationship context	Links in text only	Visual connections + documents
Layer separation	Folders or tags	Canvas layers
Agent orientation	Read file, guess context	Traverse canvas structure
Agent writes back	Edit markdown	Create objects with position and links
Human review	Read another .md	See it on canvas instantly

For AI developers

If you're building agent workflows — game design pipelines, research automation, knowledge management systems — the Levelry MCP canvas gives your agents a structured, visual workspace instead of a flat file system. Your agent's output becomes something you can see, rearrange, and build on.

For game designers

Let your AI agent draft mechanics on the canvas while you focus on creative decisions. The agent reads your existing design, understands connections between systems, and adds new nodes in the right context — all through MCP.

MCP Canvas:AI Agents on Your Knowledge Base