OpenViking
Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.
OpenViking is an open-source context database designed specifically for AI agents, unifying memory, knowledge retrieval, and skills under a virtual filesystem accessible via the viking:// protocol. It organizes agent context into a hierarchical directory structure with three-tiered loading (L0 abstract, L1 overview, L2 details) to minimize token usage while maintaining access to full information. The system provides observable, deterministic retrieval through filesystem-like commands (ls, tree, find) and integrates with Claude Code, Cursor, Codex, and other AI agents via MCP and native plugins. Benchmark results show it improves memory accuracy from 24-57% to 80-83% while reducing token consumption by 34-91%.
Key Features
Use Cases
- 01Providing long-term conversational memory for AI coding assistants like Claude Code and Cursor
- 02Building RAG knowledge bases from documentation, repositories, and web pages with efficient retrieval
- 03Storing and organizing reusable agent skills that can be discovered and invoked deterministically
- 04Debugging agent retrieval issues by inspecting the exact directory traversal path taken
- 05Reducing token costs in multi-turn agent interactions by loading only the necessary detail level
- 06Managing per-user preferences and memories in multi-agent or multi-user systems
Related MCP Servers
View moremarkitdown
Python tool for converting files and office documents to Markdown.
firecrawl
The context API to search, scrape, and interact with the web at scale. 🔥
prompts.chat
f.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source — self-host for your organization with complete privacy.
langflow
Langflow is a powerful tool for building and deploying AI-powered agents and workflows.
OpenViking — FAQ
What is OpenViking and how does it work with AI agents?+
OpenViking is a context database that stores agent memories, knowledge, and skills in a virtual filesystem under the viking:// protocol. AI agents access this context through MCP integration or native plugins, using filesystem-like commands to browse and retrieve information deterministically rather than querying opaque vector stores.
How do I install OpenViking?+
Install with pip install openviking --upgrade, then run openviking-server init to configure providers and models through an interactive wizard. Start the server with openviking-server, and use the included ov CLI to manage context. Python 3.10 or higher is required.
Which AI clients and agents does OpenViking support?+
OpenViking integrates with Claude Code, Codex, Cursor, OpenClaw, Hermes, TRAE, OpenCode, pi, and any MCP-compatible client. It also supports LangChain/LangGraph frameworks and provides Agent Plugins 1.0 compatibility.
Do I need API keys or paid services to use OpenViking?+
OpenViking itself is free and open-source under AGPLv3. However, you'll need to configure an LLM provider (Volcengine, OpenAI, Kimi, GLM, or local Ollama) for semantic processing. The init wizard helps set this up, and Ollama is a free local option.
How do I connect OpenViking to Claude Code?+
After starting the OpenViking server, follow the Claude Code integration guide at the official documentation. The integration typically involves configuring the MCP server connection in Claude Code's settings to point to your running OpenViking instance.
What are the three context layers and why do they matter?+
L0 (abstract) is a ~100-token summary for quick relevance checks, L1 (overview) provides ~2k tokens of structure and key points for planning, and L2 contains full details loaded only when needed. This tiered approach reduced token usage by 34-91% in benchmarks while maintaining or improving accuracy.
How do I install OpenViking?+
Open the source repository on GitHub and follow its README. OpenViking is a mcp server — MCP Agents Market links you directly to the official repo.
Is OpenViking free?+
OpenViking is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.