---
title: "Volcengine OpenViking: Open Context Database for AI Agents"
section: "raw"
type: "source"
created: "2026-08-23"
updated: "2026-08-23"
canonical: "https://pyweb.dev/wiki/raw/articles/volcengine-openviking-2026"
---
# Volcengine OpenViking: Open Context Database for AI Agents

**Repository:** https://github.com/volcengine/OpenViking  
**Stars:** 31k | **License:** AGPLv3 core  
**Research Paper:** *VikingMem* (arXiv:2605.29640, accepted VLDB 2026)

## System Design
- **Virtual Filesystem Context:** Unifies memories, resources, and skills under a hierarchical `viking://` URI namespace. Agents navigate context deterministically using filesystem primitives (`ls`, `tree`, `find`) rather than relying purely on vector embeddings.
- **Tiered Context Disclosure:** Structures information into L0 (Abstract), L1 (Overview), and L2 (Detailed specification) tiers to protect agent context windows.
- **Directory-Recursive Retrieval:** Preserves spatial and relational context across adjacent documents during queries.
- **Auditable Trajectories:** Leaves transparent, step-by-step retrieval logs for post-hoc error analysis and evaluation.
- **Ecosystem:** Integrated with Hermes Agent, deer-flow, and enterprise agent frameworks.

---

## Agent Navigation

### Machine endpoints
- Knowledge graph: https://pyweb.dev/api/graph.json
- Graph analysis: https://pyweb.dev/api/graph-analysis.json
- Context index: https://pyweb.dev/llms.txt