Memary vs OpenViking
Agent memory and context stores compared
Memary and OpenViking both give AI agents ways to retain and retrieve context. Memary centers on knowledge-graph retrieval and a ReAct chat agent, while OpenViking organizes documents, memories, and skills in a browsable virtual filesystem with scoped retrieval.

Memary: Add Memory to Autonomous Agents with LlamaIndex
Memary is an open-source memory layer for autonomous agents, built around LlamaIndex, knowledge graphs, and a ReAct chat agent. It stores interaction-derived entities and user context to help agents retrieve relevant information and tailor responses.

OpenViking: Organize Agent Memory, Knowledge, and Skills
OpenViking is a context database for AI agents that organizes knowledge, memories, and skills in a browsable virtual filesystem. It supports scoped retrieval and agent integrations, with options to self-host or use a hosted service.
| Memary | OpenViking | |
|---|---|---|
| Language | Jupyter Notebook | Python |
| License | MIT | AGPL-3.0 |
| Stars | 2.7k | 39.2k |
| Forks | 205 | 3.1k |
| Last analyzed | Oct 25, 2026 | Oct 3, 2026 |
Key differences
- Memary extracts entities from interactions into a knowledge graph and tracks mentions over time; OpenViking stores resources, memories, and skills under browsable `viking://` paths.
- Memary is written in Jupyter Notebook and describes a Python-agent example; OpenViking is written in Python and offers integrations, SDKs, a CLI, HTTP API, and web studio.
- Memary is MIT-licensed; OpenViking is AGPL-3.0-licensed, with component-specific and commercial self-managed licensing considerations.
- Memary describes local Ollama and OpenAI model options, and graph configuration through FalkorDB or Neo4j; OpenViking can be self-hosted or used through a hosted service.
- Memary reports no tests or CI and has had no commits for over a year; OpenViking is not archived and describes local and production deployment options.
Choose Memary if you…
- want a Python-agent starting point combining LlamaIndex knowledge-graph retrieval and interaction memory.
- need a ReAct chat-agent example with options for custom tools and separate FalkorDB graph contexts.
- prefer an MIT-licensed project and can validate installation and behavior in your environment.
Choose OpenViking if you…
- need persistent, inspectable context organized as files and directories, with retrieval scoped to a directory or subtree.
- want integrations for coding agents, MCP clients, LangChain, or Python, Go, and TypeScript SDKs.
- plan to self-host or use a hosted service and can meet the model-provider and licensing requirements.
Compare the numbers side by side, or add more projects
This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.