Hypabase

What is Hypabase?

Hypabase is a Python hypergraph library with provenance tracking and SQLite persistence. It provides an MCP server (14 tools + 2 resources), a CLI, and a Python API for storing and querying n‑ary relationships (hyperedges) where an edge can connect any number of nodes. It is intended for knowledge graphs, agent memory, and data integration scenarios where provenance (source and confidence) matters.

How to use Hypabase?

Install via uv add hypabase (or uv add hypabase[mcp] for the MCP server, uv add hypabase[cli] for CLI). Use the Python API to create a Hypabase("file.db") instance, then call .edge(), .edges(), .paths(), .node(), .nodes(), .sources(), etc. Run the MCP server with hypabase mcp. CLI commands include hypabase init, hypabase node, hypabase edge, hypabase query, and hypabase stats.

Key features of Hypabase

• Hyperedges connecting 2+ nodes in a single relationship
• Provenance tracking: every edge carries source and confidence
• SQLite persistence to a local file automatically
• O(1) vertex‑set lookup for exact node sets
• Namespace isolation via .database("name") within a single file
• MCP server with 14 tools and 2 resources for AI agents
• CLI for initialisation, nodes, edges, and queries

Use cases of Hypabase

• Representing complex real‑world relationships (e.g., medical events, academic papers) without decomposition
• Structured, queryable memory for AI agents that persists across sessions
• Storing extracted relationships for retrieval‑augmented generation (RAG)
• Biomedical data modeling (drug interactions, clinical events, molecular pathways)
• Supply chain and collaboration networks where relationships involve more than two entities

FAQ from Hypabase

What is a hypergraph and how does Hypabase differ from a regular graph database?

In a regular graph each edge connects exactly two nodes. Hypabase stores hypertedges that connect any number of nodes natively, preserving grouping without splitting into binary edges. It also tracks provenance (source and confidence) on every edge.

What are the runtime dependencies and installation requirements?

Hypabase requires Python 3.x. Install via pip (uv) – the base package pulls in SQLite (stdlib). For MCP capabilities add the mcp extra; for CLI add the cli extra.

Where does Hypabase store data?

Data is stored in a local SQLite database file specified when creating a Hypabase instance (e.g., Hypabase("my.db")). No external database server is needed.

What transport and authentication does the MCP server use?

The MCP server uses stdio transport (no network port) by default. There is no built‑in authentication; it is designed for local or trusted environments (e.g., Claude Desktop, Cursor, Windsurf).

What are the known limitations of Hypabase?

The README does not list explicit limitations. The library is a Python package and inherits SQLite single‑writer concurrency; very large graphs may require optimisation or sharding.