OpenViking is a context database for AI agents that unifies agent memory, knowledge retrieval, and skills under a single virtual filesystem interface.
The tool addresses the challenge of managing fragmented context across multiple AI agent systems by presenting memories, resources, and skills as a unified `viking://` filesystem. Agents interact with their context deterministically using familiar filesystem operations like `ls`, `tree`, and `find` rather than querying opaque vector stores. Content is automatically processed into three abstraction tiers—L0 abstract, L1 overview, and L2 details—on write and loaded on demand based on task requirements. Directory-based recursive retrieval locates the highest-scoring directory first, then drills down layer by layer to return results with surrounding context intact. Every retrieval preserves its browsing trajectory, making it possible to debug and understand exactly which path produced a given result. Sessions automatically extract user preferences and agent experience into long-term memory after commit.
Teams building AI agents that need transparent, debuggable context management should consider this tool, particularly those frustrated with black-box vector store approaches or managing multiple disconnected memory systems. The tiered loading mechanism directly reduces token consumption by loading only the abstraction depth required for a given task. The filesystem metaphor makes context manipulation deterministic and familiar to developers accustomed to file-based workflows.
The project maintains a substantial base of external adopters who report real-world issues rather than the core team driving the issue tracker. Maintainers typically respond to new issues and pull requests within a day. Work in the issue tracker centers on bug reports, user questions, and enhancement requests.
