graphiti-core

What it is and what it does

Graphiti is a temporal graph framework that builds evolving context graphs for AI agents by continuously integrating user interactions, structured and unstructured data, and external information. Unlike static knowledge graphs or traditional RAG, it tracks when facts become true and when they are superseded, maintaining full temporal history and provenance to source episodes. Each entity, relationship, and fact has a validity window, allowing queries across time, meaning, and relationships.

The framework supports incremental updates without full recomputation, hybrid retrieval combining semantic embeddings with keyword search and graph traversal, and custom entity and relationship types defined via Pydantic models. It requires a pluggable graph database backend (Neo4j, FalkorDB, or Amazon Neptune) and integrates with OpenAI for embeddings and PostHog for telemetry. The design prioritizes real-time interaction and precise historical queries over batch processing, making it suitable for applications where context must evolve with every interaction.

Use it for:

• Build AI agent memory systems that track how user preferences, relationships, and facts change over time with full audit trails.
• Query historical state of entities at any point in time without recomputing the entire graph.
• Combine semantic search with keyword matching and graph traversal to retrieve precise context for agent decision-making.
• Integrate continuously arriving structured and unstructured data into a coherent, queryable graph without batch reprocessing.
• Define custom entity and relationship types for domain-specific applications using Pydantic models.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Graphiti builds and queries temporal context graphs for AI agents, tracking how facts change over time with full provenance to source data, supporting both semantic and keyword retrieval alongside graph traversal.

Yes. Graphiti is actively maintained, has no known vulnerabilities, installs with low friction, and solves a specific problem—temporal context graphs for evolving agent memory—that traditional RAG and static knowledge graphs do not address well. The Apache-2.0 license is permissive. The main condition is that you must provision and manage a graph database backend (Neo4j, FalkorDB, or Neptune) separately; it is not a standalone solution.

Install

graphiti-core on PyPI

pip install graphiti-core

uv add graphiti-core

poetry add graphiti-core

Installing graphiti-core

Before you install

Low friction: pure Python wheel with seven runtime dependencies (neo4j, numpy, openai, posthog, pydantic, python-dotenv, tenacity). Active maintenance with recent releases and high repository engagement (29928 stars).

License in practice

Apache-2.0 permissive license allows commercial and private use with minimal restrictions — suitable for production deployments without licensing concerns.

Quickstart

pip install graphiti-core

from graphiti_core import Graphiti
from pydantic import BaseModel

class Person(BaseModel):
    name: str

graph = Graphiti()
# Add entities and facts with temporal tracking
# Query across time and relationships

Requires Python 3.10 or higher and a third-party graph database backend (Neo4j 5.26, FalkorDB 1.1.2, Amazon Neptune, or Neptune Analytics with OpenSearch).

Verify before relying

• Specific performance characteristics (sub-second latency claim) and scalability limits for production workloads.
• Detailed setup and configuration complexity for different graph database backends.
• Whether the MCP server integration mentioned in the description is included in this package or requires separate installation.

Package facts

Evidence: graphiti_core-0.29.3-py3-none-any.whl

Tags

More Artificial Intelligence packages

Further reading