Graphiti

MemoryApache-2.0

Graphiti builds and queries knowledge graphs for AI agents from their conversations and data, and updates the graph as new data comes in. It is a Python library from Zep, aimed at agents whose underlying data keeps changing over time.

The README lists Neo4j and FalkorDB setups through Docker Compose and optional extras for FalkorDB, Amazon Neptune and several model providers. The repository also contains reference implementations of an MCP server and of a REST server.

Pinout

  • MCP server: supported, verified
  • HTTP API: supported, verified
  • command line: not checked yet
  • Python SDK: not checked yet
  • TypeScript SDK: not checked yet
  • OpenAPI description: not checked yet
Where each pin comes from
InterfaceStatusSource we readRead on
MCP serverSupported, verifiedREADME, section "MCP Server" (reference implementation in mcp_server)
HTTP APISupported, verifiedREADME, section "Reference MCP and REST servers" (REST server in server)
Command lineNot checked yet——
Python SDKNot checked yet——
TypeScript SDKNot checked yet——
OpenAPI descriptionNot checked yet——

Facts

Facts about Graphiti
LicenseApache-2.0. GitHub API, repository license field, read .
Main languagePython (as detected by GitHub). GitHub API, repository record, read .
Publishergetzep (GitHub organization). GitHub API, repository record, read .
Latest GitHub releasev0.31.2, published 2026-10-20. GitHub API, latest release, read .
Graph databasesNeo4j and FalkorDB via Docker Compose; extras for FalkorDB and Amazon Neptune. README in the repository, read .
Reference serversMCP server (mcp_server directory) and REST server (server directory). README in the repository, read .
Last commit. GitHub API, latest commit on the default branch of getzep/graphiti, read .

Limits

  • The sample Docker Compose files ship with default database passwords, which the README says to change before any shared deployment.

Strengths

  • Reference MCP and REST servers live in the same repository as the library.

Alternatives to Graphiti

All memory tools for agents
  1. Open-source memory layer storing agent context across sessions and models.

    Memory
  2. Agent memory system with retain, recall and reflect operations.

    Memory
  3. Memory layer that wraps LLM clients to store and recall agent state.

    Memory
  4. Context layer for agents, built on a temporal context graph.

    Memory

Last check: . Method: public repository and docs read automatically, page and day written down. Next check due within 90 days. How we check

Is this your tool? Write to hello@agent-tools.directory from an address on its domain to correct a pin or a fact.