Open-source corporate memory platform. Extracts structured knowledge from your work tools, serves it to any AI agent.
AI agents start every session from scratch. Decisions get lost in chat history. Knowledge is scattered across a dozen tools.
Alaya runs in the background — absorbs, structures, connects. When an agent needs context, it's already there.
Sources (Slack, GitHub, Linear, docs, ...)
|
Extraction --> Entities, Claims, Relations
|
Knowledge Tree --> browsable like a book
|
CLI - MCP - API --> any AI agent
Entities — people, projects, tasks, decisions. Dynamic ontology that grows with your company.
Claims — atomic facts with provenance. "Project Alpha deadline is March 15" links back to the message where it was said. When the deadline changes, both facts are preserved with full history.
Knowledge Tree — hierarchical memory, like a table of contents. An agent reads the overview, drills into the section it needs.
Under active development. First release coming soon.
Python 3.13 / FastAPI / SQLAlchemy 2.0 + Pydantic / PostgreSQL + pgvector / Redis / TaskIQ / Anthropic SDK / Typer
BSL 1.1 — source available, self-host without restrictions. Converts to Apache 2.0 after 3 years.