Hyper-Extract CLI
A CLI and MCP server that extracts structured knowledge from documents: graphs, hypergraphs, and fact search with sources
- 4,239 ★ on the repository
- Last commit
Extended access
Why this level
- Sends document content to the LLM and embedding provider to build the graph
- Stores provider keys locally in a configuration file
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/hyper-extractDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- Sends document content to the LLM and embedding provider to build the graph
- Stores provider keys locally in a configuration file
This is third-party code. Review the repository files before installing.
What it does
Hyper-Extract parses unstructured text and builds a knowledge graph from it using one of the built-in templates, such as a biography, timeline, or report. Every fact in the graph is tied back to its source in the original document, so a claim can be traced to where it came from. The resulting knowledge base can be queried with search or RAG-style questions, and exported to Obsidian, GraphML, CSV, JSON-LD, or Cypher. A separate MCP server gives agents such as Claude Desktop read and export access to already built knowledge bases without changing their content.
Who it is for. For analysts and technical writers who need to turn a set of documents into a verifiable knowledge base.
Good fit when
- You need to build a knowledge graph from articles, reports, or documentation
- You need to quickly find facts tied back to their source inside a large document
- You need to export extracted knowledge to Obsidian or a graph database
Not a fit when
- You just need a plain text summary without graph structure
- You have no access to any LLM or embedding provider
Example request
Build a knowledge graph from this report using the biography_graph template and find all the achievements mentioned in itLimitations
Extraction and search need an LLM and embedding provider key (OpenAI, DeepSeek, Anthropic, Google Gemini, Alibaba Bailian, or a local vLLM). Some cloud providers are unavailable from Russia without a VPN, while pairing DeepSeek with any embedder works as a cheaper option.
How to disable. Remove the package: uv tool uninstall hyperextract (or pipx uninstall hyperextract) and delete the ~/.he/config.toml config file.
FAQ
How is this different from plain RAG?
Hyper-Extract first builds an explicit fact graph tied to sources, and only then can you search or ask questions over it, instead of just indexing text chunks.
Can the MCP server modify my knowledge base?
No, it is read and export only for already built knowledge bases; you cannot create or delete them through it.
Discussion
No questions or reviews yet.
Related
CLI toolsThe official Graphlit MCP server: an agent collects material from Slack, GitHub, mail and websites into a knowledge base and searches it
A local MCP server that turns a folder of documents into a knowledge base with hybrid search for Claude Code and Cursor
MCP server for the Flexible GraphRAG platform: ingest from 14 data sources, hybrid search and AI answers over a knowledge graph
A knowledge graph with an MCP server: the agent searches concepts by meaning and sees how well each is supported.
myKG builds a confidence scored knowledge graph from mixed documents and serves it to an agent through MCP or a Claude Code skill.
KGLite keeps a knowledge graph in one file, and its bundled MCP server gives agents Cypher queries with no separate database.
README in short
The README describes Hyper-Extract as a CLI that turns documents into structured knowledge with one command. The tool builds graphs and hypergraphs from templates, supports several LLM and embedding providers, and exports to multiple formats including Obsidian. A separate, read-only MCP server is documented that connects knowledge bases to Claude Desktop and other MCP clients. It is Apache-2.0 licensed and distributed via PyPI.