Skip to main content

​
MemWire Open Source Overview

MemWire is an open source & enterprise-ready AI memory infrastructure layer. MemWire follows zero-external-dependency design principle. MemWire gives your AI applications persistent, auditable memory with structured, updatable facts, fastest semantic retrieval across conversations and knowledge, integrated any document ingestion, and controlled LLM context assembly.
Truly open-source! MemWire is a fully research and community-driven project.
  • Fully customizable — adapt schemas, memory categories, and features to your use case
  • Self-hosted — run entirely on your local machine, on-premise or in your own cloud
  • Multi-tenant — isolate applications, users, and workspaces securely
  • Your database — Use your own database: PostgreSQL pgvector, Qdrant, Pinecone, ChromaDB, Weawiate or your preferred stack
  • Your LLM — Use your own LLM: OpenAI, Anthropic, Gemini, Ollama, or any provider
  • Deploy anywhere — edge, private cloud, public cloud (AWS, Microsoft Azure, Google Cloud), air-gapped environments
  • Knowledge ingestion — ingest documents (PDF, Excel, CSV, etc.) alongside conversation memory, recalled together at query time
  • Auditable — every memory is traceable, categorized (fact, preference, instruction, event, entity), and inspectable

​
Why MemWire?

Fully customizable

Adapt schemas, memory types, and storage to your use case. Swap the embedding model, tune recall sensitivity, or add your own memory categories.

Self-hosted

Run entirely on your local machine, on-premise or in your own cloud. Your data never leaves your infrastructure.

Multi-tenant

Isolate applications, users, and workspaces securely out of the box.

Model-agnostic

Bring your own LLM — OpenAI, Anthropic, Gemini, Ollama, or any provider.

Graph-powered recall

Memory is stored as a displacement graph and retrieved very fast.

Zero-LLM classification

Memories are categorized using anchor-based classification and no LLM call required. Instant, cost-free tagging with zero added latency.

​
Get started

Quickstart

Add persistent memory to your AI agent in minutes.

API reference

Explore the REST API endpoints for memory store, recall, and search.