volcengine/OpenViking

Self-evolving Context Database for AI Agents. Unify Agent Memory, Knowledge RAG and Skills.

What it solves

OpenViking provides a transparent, structured way for AI agents to store and retrieve knowledge, memories, and skills. Unlike traditional vector databases that act as "black boxes," OpenViking organizes context as a virtual filesystem, allowing both agents and human users to inspect, edit, and manage what an agent knows.

How it works

It implements a virtual filesystem under the viking:// URI scheme. Agents interact with this system using familiar file-like commands (such as ls, read, and grep). To optimize retrieval, it uses a three-tier loading system:

  • L0 (Abstract): A one-sentence summary for quick relevance checks.
  • L1 (Overview): Core information and usage scenarios for planning.
  • L2 (Details): The full original data, loaded only when necessary.

Semantic search can be scoped to specific directories or subtrees rather than scanning a flat index, and sessions can be archived as Markdown files for later inspection or merging.

Who it’s for

  • AI Agent Developers: Those building agents that require long-term memory and structured knowledge bases.
  • AI Agent Users: Users of tools like Claude Code, Cursor, or TRAE who want cross-session memory and the ability to audit their agent's knowledge.

Highlights

  • Virtual Filesystem Interface: Treats agent context as a directory structure, making it human-readable and manageable.
  • Multi-tier Context: Reduces token usage and latency by allowing agents to read summaries before full content.
  • L0/L1/L2 Layering: Specifically designed to reduce input tokens and query latency.
  • Broad Integration Support: Plugins available for major coding agents (Claude Code, Cursor, TRAE) and frameworks like LangChain.
  • Open Source: Available under the AGPLv3 license.

Related

  • Project
    vercel-labs/knowledge-agent-template

    An open-source template for building AI agents that search knowledge bases using file-system commands instead of vector databases, deployable across web, GitHub, and Discord.

  • Project
    VectifyAI/OpenKB

    An open-source system that compiles raw documents into a structured, interlinked wiki-style knowledge base using LLMs and vectorless, reasoning-based retrieval.

  • Project
    memodb-io/Acontext

    Acontext is an open-source memory layer that captures AI agent learnings as human-readable Markdown skill files, making agent memory transparent, portable, and easy to debug.

  • Project
    beilusaiying/always-accompany

    A multi-purpose AI agent framework focused on context and attention mechanisms, providing layered memory and dynamic retrieval for long-term companionship, coding, and work.

  • Project
    thecodacus/understory

    understory is an open‑source, markdown‑based knowledge‑graph memory layer for LLM agents. It provides a deterministic OKF‑compliant bundle, a server exposing tool‑calling endpoints (memory_query, memory_add, etc.), a React UI with a force‑directed graph, CLI utilities, and multi‑profile isolation. Works with any OpenAI‑ or Anthropic‑compatible model (including local llama.cpp) and runs via Docker or from source.