akillness graphiti
Graphiti is a leading agent memory in the 2026 AI agent engineering landscape (28.3k★).
When to use this skill
- Real-time temporal knowledge graphs for agent memory
- Episodes (text/message/json) incrementally ingested — no batch recompute
- Bi-temporal model: event time + ingest time, edge invalidation on contradiction
- Hybrid retrieval: semantic + BM25 + graph traversal, with rerankers
- Custom entity/edge types via Pydantic models
- Backends: Neo4j, FalkorDB (+embedded falkordblite), Kuzu, Amazon Neptune; MCP server included
- Ecosystem overlap: agent-memory-architecture references Zep/Graphiti conceptually; this covers building on graphiti-core directly
Installation
pip install graphiti-core # extras: [falkordb], [kuzu], [neptune], [anthropic], [google-genai]...
Core API
from graphiti_core import Graphiti
graphiti = Graphiti("bolt://localhost:7687", "neo4j", "password")
await graphiti.build_indices_and_constraints()
await graphiti.add_episode(name="conv-1", episode_body="Alice moved to Paris", source_description="chat")
results = await graphiti.search("Where does Alice live?")
Instructions
Step 1: Initialize the environment
Verify you have Python installed and set up properly. Use the correct package manager (pip, npm, pnpm, uv) as specified in the installation section.
Step 2: Establish the agent flow
Architecture pattern: Graph-as-memory service; agents write episodes and query facts; communities cluster entities; powers Zep's hosted memory
Step 3: Implement tool integrations and policies
Integrate necessary tools, environment keys, and safety gates according to your architectural needs. Position in the ecosystem: Temporal-KG lane of memory; alternatives: mem0 (extraction pipeline), plain RAG (no time model); complements agent-memory-architecture's temporal-vs-snapshot tradeoff
Examples
Example 1: Minimal Agent Run
from graphiti_core import Graphiti
graphiti = Graphiti("bolt://localhost:7687", "neo4j", "password")
await graphiti.build_indices_and_constraints()
await graphiti.add_episode(name="conv-1", episode_body="Alice moved to Paris", source_description="chat")
results = await graphiti.search("Where does Alice live?")
Best practices
- Follow the primary installation commands for Graphiti to ensure dependencies are resolved correctly.
- Configure model providers and API keys securely before starting the runtime.
- Use appropriate sandboxing or permission isolation if executing AI-generated code.