Question 1

What is OpenZync?

Accepted Answer

OpenZync is an open-source, persistent memory infrastructure for AI agents. It stores agent interactions as a knowledge graph — entities, relationships, and facts — that agents can read from, write to, and reason over across sessions.

Question 2

How is OpenZync different from a regular database?

Accepted Answer

Unlike a generic database, OpenZync is purpose-built for agent memory. It provides graph-based storage optimized for agentic access patterns, built-in support for multiple graph backends, LLM-integrated retrieval, webhook-driven events, and observability — all behind a single, clean API.

Question 3

Can I attach files to memory?

Accepted Answer

Yes. Files attach to session ingestion as blobs — per-blob size is org-configurable with a 50MB default, while message text is capped at 64KB. Up to 10 blobs per message and 50 per request. Each blob is stored S3-backed and referenced by messages via blob_id, and extracted content feeds the fact store, where later corrections supersede earlier facts.

Question 4

What graph backends does OpenZync support?

Accepted Answer

OpenZync supports PostgreSQL (built-in, no external graph DB required), FalkorDB, and SurrealDB. The backend is swappable via configuration — your agent memory isn't locked into any single vendor.

Question 5

Which LLM providers can I use with OpenZync?

Accepted Answer

OpenZync works with OpenAI, Anthropic (Claude), Azure OpenAI, OpenRouter, and local models through Ollama. You can route different tasks to different providers based on cost, latency, or capability requirements.

Question 6

Is OpenZync open source?

Accepted Answer

Yes, OpenZync core is open source under the AGPL v3 license (dual-license with commercial option for SaaS use). The Python SDK is Apache 2.0. You can self-host, modify, and extend it freely.

Question 7

Can I use OpenZync for free?

Accepted Answer

OpenZync is free and open source for self-hosted deployments. It's AGPL v3 licensed with a commercial option for SaaS use. Self-host without restrictions and only require a commercial license if you offer it as a hosted service without releasing modifications.

Question 8

How does human-in-the-loop work?

Accepted Answer

Human-in-the-loop approval gates are on the roadmap. The webhook system and event-driven architecture provide the foundation, and we plan to add configurable approval workflows in a future release.

Question 9

What languages/frameworks does OpenZync support?

Accepted Answer

OpenZync provides a Python SDK as a first-class citizen, plus a REST API that works with any language. A LangChain integration is also available. The SDK and API are language-agnostic via HTTP.

Question 10

Can I self-host OpenZync?

Accepted Answer

Yes. OpenZync is designed for self-hosting from day one. You can deploy it on your own infrastructure using Docker Compose or Kubernetes (Helm charts provided). We provide Docker images, Helm charts, and reference infrastructure configs for production deployments.

Question 11

How do I migrate from another memory solution?

Accepted Answer

The REST API makes it straightforward to export data from your existing system and import it into OpenZync. The underlying PostgreSQL schema is well-documented for direct data access. Migration tools for common sources are on the roadmap.

Question 12

What are the API rate limits?

Accepted Answer

Self-hosted deployments have no rate limits imposed by the platform — you control your own infrastructure. The platform includes configurable rate limiting middleware that you can tune for your deployment.

Question 13

Is my data portable?

Accepted Answer

Absolutely. OpenZync uses standard PostgreSQL tables for the default graph backend. You can export your data as JSON via the REST API or access the underlying database directly. Your data is never locked in.

Question 14

How does OpenZync handle production scaling?

Accepted Answer

OpenZync is built for production from the ground up. It supports horizontal scaling through database connection pooling, Redis caching, and async worker processing. Self-hosted deployments can be scaled using standard infrastructure patterns like horizontal pod autoscaling and database read replicas.

Frequently asked questions

Run the memory layer on your own infrastructure.