How is Cognee different from standard RAG?

Standard RAG splits documents into chunks and finds similar text using vector search — but it loses the relationships between facts. Cognee builds a knowledge graph that preserves entity relationships, temporal connections, and hierarchical structures. When you search, Cognee uses graph traversal combined with vector similarity and LLM reasoning, resulting in more accurate, context-aware answers that understand HOW facts relate to each other.

What search types are available?

Cognee supports four retrieval strategies: GRAPH_COMPLETION (default — combines vector search + graph traversal + LLM reasoning for context-aware answers), SUMMARIES (fast hierarchical overview search), INSIGHTS (structured entity relationships), and CHUNKS (pure vector similarity for raw text passages). Each strategy optimizes for different use cases.

Is Cognee open-source?

Yes! Cognee is fully open-source under the Apache 2.0 license. You can self-host the entire platform including the knowledge graph engine, vector database, and API server. A managed cloud version with API keys is also available for teams that prefer not to manage infrastructure.

Cognee MCP Server for Pydantic AI 4 tools — connect in under 2 minutes

Name: Cognee
Availability: InStock
Author: Vinkius

ai.pydantic.dev/mcp

Built by Vinkius GDPR 4 Tools SDK

Pydantic AI brings type-safe agent development to Python with first-class MCP support. Connect Cognee through Vinkius and every tool is automatically validated against Pydantic schemas. catch errors at build time, not in production.

Get MCP Server for AI Agents

ASK AI ABOUT THIS MCP SERVER

Open in ChatGPT Open in Claude Open in Perplexity

Vinkius supports streamable HTTP and SSE.

python

import asyncio
from pydantic_ai import Agent
from pydantic_ai.mcp import MCPServerHTTP

async def main():
    # Your Vinkius token. get it at cloud.vinkius.com
    server = MCPServerHTTP(url="https://edge.vinkius.com/[YOUR_TOKEN_HERE]/mcp")

    agent = Agent(
        model="openai:gpt-4o",
        mcp_servers=[server],
        system_prompt=(
            "You are an assistant with access to Cognee "
            "(4 tools)."
        ),
    )

    result = await agent.run(
        "What tools are available in Cognee?"
    )
    print(result.data)

asyncio.run(main())

Fully ManagedVinkius Servers

60%Token savings

High SecurityEnterprise-grade

IAMAccess control

EU AI ActCompliant

DLPData protection

V8 IsolateSandboxed

Ed25519Audit chain

<40msKill switch

Stream every event to Splunk, Datadog, or your own webhook in real-time

* Every MCP server runs on Vinkius-managed infrastructure inside AWS - a purpose-built runtime with per-request V8 isolates, Ed25519 signed audit chains, and sub-40ms cold starts optimized for native MCP execution. See our infrastructure

About Cognee MCP Server

Connect your AI agent to Cognee — the open-source knowledge graph platform that transforms unstructured data into structured, searchable knowledge.

Pydantic AI validates every Cognee tool response against typed schemas, catching data inconsistencies at build time. Connect 4 tools through Vinkius and switch between OpenAI, Anthropic, or Gemini without changing your integration code. full type safety, structured output guarantees, and dependency injection for testable agents.

What you can do

Add Data — Ingest raw text, documents, or structured data into named datasets. Cognee processes and stores the data for subsequent graph construction
Cognify — Transform ingested data into a structured knowledge graph by automatically extracting entities, relationships, and semantic connections
Search Knowledge — Query the knowledge graph using four retrieval strategies: graph-aware completion (LLM + graph traversal), summaries, structured insights, or raw vector similarity
Get Insights — Retrieve structured entity relationships showing how concepts connect across your knowledge base

Why Cognee over standard RAG?

Relationship-aware — understands HOW facts connect, not just what they say
Graph + Vector hybrid — combines graph traversal with semantic search for superior recall
Temporal awareness — tracks when facts were added and reason over time-based connections

The Cognee MCP Server exposes 4 tools through the Vinkius. Connect it to Pydantic AI in under two minutes — no API keys to rotate, no infrastructure to provision, no vendor lock-in. Your configuration, your data, your control.

How to Connect Cognee to Pydantic AI via MCP

Follow these steps to integrate the Cognee MCP Server with Pydantic AI.

Install Pydantic AI

Run pip install pydantic-ai

Replace the token

Replace [YOUR_TOKEN_HERE] with your Vinkius token

Run the agent

Save to agent.py and run: python agent.py

Explore tools

The agent discovers 4 tools from Cognee with type-safe schemas

Why Use Pydantic AI with the Cognee MCP Server

Pydantic AI provides unique advantages when paired with Cognee through the Model Context Protocol.

Full type safety: every MCP tool response is validated against Pydantic models, catching data inconsistencies before they reach your application

Model-agnostic architecture. switch between OpenAI, Anthropic, or Gemini without changing your Cognee integration code

Structured output guarantee: Pydantic AI ensures tool results conform to defined schemas, eliminating runtime type errors

Dependency injection system cleanly separates your Cognee connection logic from agent behavior for testable, maintainable code

Cognee + Pydantic AI Use Cases

Practical scenarios where Pydantic AI combined with the Cognee MCP Server delivers measurable value.

Type-safe data pipelines: query Cognee with guaranteed response schemas, feeding validated data into downstream processing

API orchestration: chain multiple Cognee tool calls with Pydantic validation at each step to ensure data integrity end-to-end

Production monitoring: build validated alert agents that query Cognee and output structured, schema-compliant notifications

Testing and QA: use Pydantic AI's dependency injection to mock Cognee responses and write comprehensive agent tests

Cognee MCP Tools for Pydantic AI (4)

These 4 tools become available when you connect Cognee to Pydantic AI via MCP:

cognee_add_data

After ingestion, use the cognify tool to process the data into a structured knowledge graph with entities and relationships. Ingest text or documents into the Cognee knowledge base. This is the first step before building a knowledge graph

cognee_cognify

This step extracts entities, identifies relationships, generates embeddings, and creates the graph structure needed for intelligent search. Process ingested data into a structured knowledge graph. Extracts entities, relationships, and builds a searchable graph structure

cognee_get_insights

Useful for understanding relationships between topics, discovering hidden connections, and building comprehensive knowledge views. Retrieve structured entity relationships and insights from the knowledge graph

cognee_search

Search the knowledge graph using natural language. Returns context-aware answers using graph traversal and semantic search

Example Prompts for Cognee in Pydantic AI

Ready-to-use prompts you can give your Pydantic AI agent to start working with Cognee immediately.

"Add this research data to my knowledge base: 'Transformer models were introduced by Vaswani et al. in 2017 in the paper Attention Is All You Need. They use self-attention mechanisms and have become the foundation for models like GPT, BERT, and T5.'"

"Process my data into a knowledge graph."

"What is the relationship between Transformers and GPT?"

Troubleshooting Cognee MCP Server with Pydantic AI

Common issues when connecting Cognee to Pydantic AI through the Vinkius, and how to resolve them.

MCPServerHTTP not found

Update: pip install --upgrade pydantic-ai

Cognee + Pydantic AI FAQ

Common questions about integrating Cognee MCP Server with Pydantic AI.

How does Pydantic AI discover MCP tools?

Create an MCPServerHTTP instance with the server URL. Pydantic AI connects, discovers all tools, and generates typed Python interfaces automatically.

Does Pydantic AI validate MCP tool responses?

Yes. When you define result types as Pydantic models, every tool response is validated against the schema. Invalid data raises a clear error instead of silently corrupting your pipeline.

Can I switch LLM providers without changing MCP code?

Absolutely. Pydantic AI abstracts the model layer. your Cognee MCP integration works identically with OpenAI, Anthropic, Google, or any supported provider.