KnowMapped includes built-in AI generation for turning documents, text, and prompts into knowledge maps. MCP extends that capability outside the application by allowing compatible AI assistants to work with KnowMapped during a conversation.
MCP, or Model Context Protocol, is an open standard for connecting AI applications with external tools and data sources. Rather than requiring a user to move information manually between an assistant and KnowMapped, an MCP-enabled client can call KnowMapped directly when creating or modifying a map.
For developers, this means KnowMapped can become part of a broader AI workflow rather than functioning only as a destination for imported information.
Built-in generation and MCP
KnowMapped supports two related AI workflows:
-
Built-in generation in KnowMapped. A user can upload a document, provide text, or describe a topic, and KnowMapped generates an editable map from that source material.
-
External AI applications connected through MCP. The assistant works with KnowMapped through the protocol while the user remains in the conversational environment.
Both approaches create editable KnowMapped maps. The difference is when mapping enters the workflow: built-in generation organizes source material provided to KnowMapped, while MCP allows an external AI application to create or modify a map as analysis develops.
Once connected, an AI assistant can create maps, add concepts and relationships, and continue working with existing maps. Those maps persist independently of the conversation and can be opened, edited, shared, and developed further in KnowMapped.
This is useful when relationships among concepts carry much of the meaning, including system architecture, research synthesis, project dependencies, policy analysis, meeting outcomes, and conceptual models.
A researcher might map concepts emerging from a discussion of several sources. A development team might update a system map as components and dependencies are identified.
The Model Context Protocol specification and documentation are available at modelcontextprotocol.io.
Connecting to KnowMapped MCP
KnowMapped MCP is available to Professional users.
The server endpoint is:
https://mcp.knowmapped.com/mcp
The service uses HTTP streaming and authenticated access. A compatible client connects to the endpoint, completes authorization with KnowMapped, and can then request the mapping operations exposed to that client.
Setup varies by application. ChatGPT, Codex, and other MCP-compatible clients each provide their own interface for adding MCP servers and authorizing connections.
Creating a new map through MCP consumes one KnowMapped generation credit. Working with an existing map through MCP does not consume generation credits.
Once the connection is active, users can ask the assistant to create a map of the topic under discussion or modify an existing map as new concepts and relationships emerge.
From a chat to a map
The example in the above image began with a short request to create a knowledge map of the Lake Superior ecosystem.
The user did not define nodes, provide an outline, or specify individual relationships. The AI assistant interpreted the request, its context, identified relevant concepts and connections, and created the initial map through KnowMapped MCP.
The resulting map remains an interpretation of the available information and should be reviewed like any AI-generated output. Working in the KnowMapped app or their connect AI tool, users can revise concepts, remove weak connections, add missing information, and reorganize the structure.
See the MCP integration guide for connection and authorization details.