OpenViking
An open-source context database for AI agents that unifies memory, resources, and skills management through a filesystem paradigm with tiered context loading and self-evolving memory.
At a Glance
About OpenViking
OpenViking is an open-source Context Database designed specifically for AI Agents, built by Volcengine. It abandons the fragmented vector storage model of traditional RAG and adopts a filesystem paradigm to unify the structured organization of memories, resources, and skills needed by agents. With tiered context loading (L0/L1/L2), directory recursive retrieval, and automatic session management, OpenViking dramatically reduces token consumption while improving retrieval accuracy and enabling agents to self-evolve over time.
- Filesystem Management Paradigm: Unifies memories, resources, and skills under a
viking://virtual filesystem, enabling deterministic context manipulation via standard commands likelsandfind. - Tiered Context Loading (L0/L1/L2): Automatically processes context into abstract, overview, and full-detail layers loaded on demand, significantly reducing token costs.
- Directory Recursive Retrieval: Combines intent analysis, vector retrieval, and recursive directory exploration to achieve precise, globally-aware context retrieval.
- Visualized Retrieval Trajectory: Exposes the full retrieval path for every query, making context debugging transparent and actionable.
- Automatic Session Management: Extracts long-term memory from sessions asynchronously, updating user preferences and agent experience so agents get smarter with use.
- VikingBot Agent Framework: Built-in AI agent framework (
openviking[bot]) with interactive chat support, installable via pip. - Multi-Provider VLM & Embedding Support: Works with Volcengine (Doubao), OpenAI, OpenAI Codex, Kimi, GLM, Google Gemini, Ollama, and many more providers.
- Interactive Setup Wizard:
openviking-server initauto-detects hardware, recommends models, and generates a ready-to-use configuration;openviking-server doctorvalidates the full setup. - Cross-Platform CLI: Rust-based
ovCLI for managing resources, running semantic search, and interacting with the server on Linux, macOS, and Windows. - OpenClaw & OpenCode Plugin Support: Provides ready-made memory plugin examples for OpenClaw, OpenCode, and Claude Code agents.
Community Discussions
Be the first to start a conversation about OpenViking
Share your experience with OpenViking, ask questions, or help others learn from your insights.
Pricing
Open Source
Fully free and open-source under AGPLv3. Self-host on any supported OS.
- Full context database functionality
- Filesystem management paradigm
- Tiered context loading (L0/L1/L2)
- Directory recursive retrieval
- Visualized retrieval trajectory
Capabilities
Key Features
- Filesystem-paradigm context management (viking:// URIs)
- L0/L1/L2 tiered context loading
- Directory recursive retrieval with intent analysis
- Visualized retrieval trajectory
- Automatic session memory extraction and self-evolution
- VikingBot AI agent framework
- Multi-provider VLM support (Volcengine, OpenAI, Codex, Kimi, GLM, Gemini, Ollama)
- Multi-provider embedding support (Volcengine, OpenAI, Azure, Jina, Voyage, Cohere, etc.)
- Interactive setup wizard (openviking-server init)
- Health check command (openviking-server doctor)
- Rust-based CLI (ov)
- OpenClaw, OpenCode, and Claude Code memory plugin examples
- Docker image with bundled VikingBot
- AGPLv3 open-source license