# AgensGraph 2.18 adds ISO GQL support and performance improvements via PostgreSQL 18.6

> **Key takeaway:** AgensGraph 2.18 introduces initial ISO GQL standard clauses and significant traversal performance gains by building on PostgreSQL 18.6.

**Published:** 2026-11-07T14:02:28.141094+00:00 
**Source:** PostgreSQL Official News 
**Category:** databases-storage 
**Canonical URL:** https://fosswire.org/news/agensgraph-218-adds-iso-gql-support-and-performance-improvem.html 

## Summary
The AgensGraph development team has released version 2.18.6.1, a major update to the open-source multi-model graph database. Built on top of PostgreSQL 18.6, this release focuses on aligning the engine with the ISO/IEC 39075 (GQL) international standard for property graphs while significantly optimizing query execution speeds.

Key improvements target both the query language and the underlying storage engine. The release introduces new GQL clauses alongside existing openCypher support and implements indexed edge label lookups for variable-length traversals. Additionally, the update includes a major overhaul of the Python driver and new integrations for AI-driven GraphRAG applications.

## Technical improvements in AgensGraph 2.18
- **Query Language Expansion:** Adds initial ISO GQL clauses including LET, FOR, FILTER, NEXT, FINISH, and INSERT, as well as graph subqueries such as CALL { }, EXISTS { }, and COLLECT { }.
- **Traversal Optimization:** Variable-length traversals now utilize edge label indexes at every hop rather than scanning the entire label, reducing two-hop traversal times from over 20 seconds to milliseconds on large datasets.
- **PostgreSQL 18.6 Integration:** Inherits core features including asynchronous I/O, uuidv7() support, virtual generated columns, OAuth authentication, and B-tree skip scans.
- **Python Driver and AI Tooling:** The agensgraph-python 2.1 driver is rebuilt on psycopg 3 with async support and direct loading into pandas, Polars, or Arrow; agensgraph-ai provides integrations for LangChain, LlamaIndex, and Model Context Protocol (MCP) servers.
- **Security and Storage:** Implements row-level security (RLS) enforcement for Cypher writes and fixed-length patterns, and introduces typed property columns for faster filtering and sorting via native PostgreSQL indexing.

## Why it matters
By moving toward the ISO GQL standard, AgensGraph provides a more standardized path for developers transitioning from traditional relational databases to graph models. The integration of pgvector and new AI frameworks allows engineers to build complex GraphRAG systems within a single, unified PostgreSQL-based environment.