# OpenViking

OpenViking is an open source AI Agent context database initiated and maintained by the ByteDance Volcano Engine Viking team for agent developers and AI application teams. It manages knowledge, long-term memory, and skills in a unified way, helping to solve the problems of agents losing context across sessions, dispersion of data, and difficulty in reusing work experience. Users can import documents, web pages, or code repositories, and then plug OpenViking into an existing agent. The system organizes the content into browsable directories and generates summaries and overviews, allowing the agent to read the required information according to the task. It provides both retrieval capabilities and support for viewing and editing the content saved by the agent, facilitating the continuous maintenance of knowledge and memory. Unified context management of core functions: Put documents, code materials, user memories and skills into the same directory structure, support browsing, searching and reading, and facilitate the management of information available to the agent. Layered on-demand reading: Generate a summary, overview, and retain the original text for the content. The agent can judge the relevance first, and then read the details to reduce loading a large amount of data at once. Long-term memory across sessions: After the session is submitted, user preferences, events, and work experience are extracted based on the content of the conversation to update the existing memory and preserve a useful background for subsequent tasks. Existing agent access: support access to Claude Code, Codex, Cursor, etc.

OpenViking is an open source AI Agent context database initiated and maintained by the ByteDance Volcano Engine Viking team for agent developers and AI application teams. It manages knowledge, long-term memory, and skills in a unified way, helping to solve the problems of agents losing context across sessions, dispersion of data, and difficulty in reusing work experience. Users can import documents, web pages, or code repositories, and then plug OpenViking into an existing agent. The system organizes the content into browsable directories and generates summaries and overviews, allowing the agent to read the required information according to the task. It provides both retrieval capabilities and support for viewing and editing the content saved by the agent, facilitating the continuous maintenance of knowledge and memory. Unified context management of core functions: Put documents, code materials, user memories and skills into the same directory structure, support browsing, searching and reading, and facilitate the management of information available to the agent. Layered on-demand reading: Generate a summary, overview, and retain the original text for the content. The agent can judge the relevance first, and then read the details to reduce loading a large amount of data at once. Long-term memory across sessions: After the session is submitted, user preferences, events, and work experience are extracted based on the content of the conversation to update the existing memory and preserve a useful background for subsequent tasks. Existing agent access: support access to Claude Code, Codex, Cursor, etc.

Canonical: https://uniquebase.ai/products/openviking
Stable identifier: UB-D6614
Snapshot captured: 2026-10-28T00:47:51.550Z
English updated: 2026-10-28T00:56:06.605Z

## Connected profiles

- [OpenViking](/companies/openviking)

## Sources

- No source links are available for this profile.