go-bibx

go-bibx

go-bibx builds and analyses citation networks with the Tree of Science (ToS) method. It fetches works from OpenAlex, stores them in a local SQLite database, and works as a command-line tool, an MCP server for AI agents, and a Go library. Maintained by Core of Science.

The Tree of Science

The SAP algorithm uses the topology of the citation graph to map how a research field evolved. It sorts works into:

• Roots: Seminal works that many citation chains in the collection end at as their original source.
• Trunk: Structural works where the topic took its current shape, joining two or more roots.
• Leaves: Current, state-of-the-art works whose citation chains lead back to the roots.

Roots can look a bit off topic. Topics seldom start under their conventional modern names, so the works they grew from may use other terms or come from neighbouring fields: for "apache kafka", roots include ZooKeeper and MapReduce. Judge a root by whether the topic grew from it, not by whether it mentions the query. Generic method citations, such as reporting guidelines, are the exception: many papers cite them, but the topic did not grow from them.

Installation

Download a binary from GitHub Releases, or install with Go 1.27+:

go install github.com/coreofscience/go-bibx/cmd/bibx@latest

CLI Usage

# Fetch works on a topic, run the analysis, and save a named collection
bibx inquire "quantum machine learning" --limit 300 --name qml

# Top works by category in a collection (defaults to the latest)
bibx query root --limit 10 --collection qml --format reference

# Keyword search, returning the matches and the citation paths connecting them
bibx search "entanglement" --collection qml

# Explore the citation graph in the browser at http://localhost:8080
bibx view --collection qml

# Export a collection as BibTeX
bibx collection export qml --output qml.bib

# Manage collections, by name or UUID
bibx collection list
bibx collection show qml
bibx collection delete qml

inquire matches every term of the query against titles and abstracts, so use one concept of 1-3 words per call. Results are cached per query and limit; pass --force to fetch them again.

Collections live in .bibx/bibx.db under the current directory. Use the global flags --root or --db, placed before the subcommand, to change that. Run bibx <command> --help for all the options and output formats.

MCP Server

bibx mcp serves the same collections to AI agents over stdio, using the Model Context Protocol. Add it to your client's configuration:

{
  "mcpServers": {
    "bibx": {
      "command": "bibx",
      "args": ["--db", "/absolute/path/to/bibx.db", "mcp"]
    }
  }
}

Pass an absolute --db path, since clients may start the server from any directory.

Library Usage

Import the public packages to build on go-bibx: services for the domain operations, repository for the database, clients/openalex for the OpenAlex API, algorithms for the graph algorithms, models for the shared types, and mcp to serve the MCP server on your own transport. See the package documentation.

Development

See CONTRIBUTING.md for the workflow and docs/adrs for the architecture decisions.

License

This project is licensed under the MIT License.