- STRUCTURE
- REASONING
- RETRIEVAL
- HUMAN
Short answer
How can I turn my text files into a knowledge graph?
You feed your raw documents into the MCP, and the agent extracts entities and their relationships to build a structured graph. This gives your agent a map of how different concepts in your files actually relate to one another.
The results of a structured graph.
Where your data lands.
Once the agent processes your files, you get a structured map of information.
STRUCTURE
Entity networks
The agent identifies key people, places, and concepts. It turns a flat pile of text into a web of interconnected data.
REASONING
Logical connections
Because relationships are mapped, your agent can follow paths between distant ideas. This makes complex reasoning much more accurate.
RETRIEVAL
Contextual search
Instead of just matching keywords, your agent understands the context of a query. It finds information based on how things are linked.
HUMAN
Manual review
You can inspect the extracted relationships to ensure the logic holds up. You decide which connections are meaningful enough to keep.
The workflow
What your AI does when the data arrives.
The agent moves from raw text to a navigable map of knowledge.
Ingest text
The agent takes your unstructured files and prepares them for processing.
cognee_add_dataExtract entities
The agent scans the text to find specific subjects and objects.
cognee_cognifyMap relationships
The agent identifies how those entities interact or relate to each other.
cognee_cognifyFinalize graph
The system completes the structured network for your agent to use.
cognee_cognify
Try it
Copy these to start.
Use these prompts to kick off the graph building process.
Starting points
Swap the bracketed text for your specific file paths or folder names.
Cognee Connector
You're all set. Choose your MCP client and follow the setup instructions.
https://edge.vinkius.com/vk_preview_c6i3hXOYfoGrkpkYTLcBWkNZG3FKUa2WjFOiOwHJ/mcpClaude Desktop
Follow the steps below to connect in seconds.
- 1In Claude Desktop, open Settings → Connectors.
- 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
- 3Click Add and start a new chat — Cognee capabilities are ready to use.
{
"mcpServers": {
"cognee-mcp": {
"url": "https://edge.vinkius.com/vk_preview_c6i3hXOYfoGrkpkYTLcBWkNZG3FKUa2WjFOiOwHJ/mcp"
}
}
}Add these research papers to my knowledge base: [path/to/folder]
Cognify the data I just added from [filename.txt] so I can ask questions about it.
Look at the entities in my project notes and tell me how [Person A] is connected to [Project B].
Build a knowledge graph from my legal documents in [directory] and find all conflicting clauses.
Claude
ChatGPT
Cursor
VS Code
Windsurf
Claude Code
JetBrains
Cline
Start here
Connect Cognee once, then ask.
One click connects the MCP to your client. Your credentials stay encrypted, and you can start building graphs immediately.
Connect Cognee to your AIFAQ
How the graph building behaves.
- 01
Can the AI delete my original files?
No. The tools only read and process your data to build the graph. They cannot delete or modify your files.
- 02
Does the AI decide which relationships are important?
The agent proposes the connections it finds, but you decide which parts of the graph are useful.
- 03
Can I use this with any text file?
Yes, as long as the text is unstructured and readable, the agent can process it into entities.
- 04
Will this replace my existing database?
No. This creates a structured layer on top of your documents to help your agent understand them better.
- 05
Can the AI edit my documents to add the graph data?
No. The agent only reads the data to build the graph; it does not write back to your source files.
More questions about Cognee? The Connector page answers them. See everything the Cognee Connector can do
Connect Cognee to Claude, Cursor, ChatGPT & more
