ayushagrawal288/memex

Overview

Memex is a production-grade persistent memory service for AI agents built with FastAPI, pgvector, and local ONNX embeddings. It provides a REST API for storing and retrieving vector embeddings, enabling agents to maintain long-term context across sessions.

Best for

Best for
Developers building AI agents that need a self-hosted, scalable memory layer

Use cases

• Give AI agents persistent memory across conversations or tasks
• Store and query vector embeddings for semantic search in agent workflows
• Build a scalable memory backend for multi-agent systems

How to use

Install

docker compose up

Tools exposed

• store_memory
• search_memories
• delete_memory
• count_memories

Notes

Memex is a production-grade persistent memory service for AI agents built with FastAPI, pgvector, and local ONNX embeddings. It provides a REST API for storing and retrieving vector embeddings, enabling agents to maintain long-term context across sessions.

0 stars on GitHub. Last updated 2026-06-04. Licensed MIT.

Use cases

• Give AI agents persistent memory across conversations or tasks
• Store and query vector embeddings for semantic search in agent workflows
• Build a scalable memory backend for multi-agent systems

Pros

• Uses pgvector for efficient vector storage and retrieval
• Runs locally with ONNX embeddings, no external API dependencies
• FastAPI provides a modern, performant REST interface

Cons

• Requires PostgreSQL with pgvector extension, adding infrastructure complexity
• Limited documentation and community support due to zero stars
• ONNX embeddings may have lower accuracy than cloud-based alternatives

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.