Skip to content
Tooling
Index your codebases

Index your codebases

Codebases might sometimes hold a lot of files, with various structures, interactions and architectures. When an agent searches something inside of a codebase, its primary action is to either read all files it needs when searching (blindly) or run commands (e.g. grep)s to filter files based on their content.

It’s even more visible when searching code dependencies, what calls this function, what uses this class or struct, etc.

To fix this behavior, codebases can be indexed through graph-based databases instead. By doing this, an agent can use an MCP server, a CLI, or a skill to easily retrieve all interactions regarding a class, a function, a method, etc. with a single request.

And it’s not just language-based codebases that can be indexed: it can also be used for Helm charts, Terraform modules, markdown references, etc. (still depends on each tool’s feature set).

Warning

Present in Boost documentation but absent from the CLI (as of 2026-09-25) and from the locally deployed UI

  • Upstream: https://github.com/jfrog/boost
  • Indexing: through the CLI or automatically via a background hook on session start
  • Navigation: through the MCP server

Install:

mise use -g 'github:jfrog/boost@latest'
# or
curl -fsSL https://boost.jfrog.com/install.sh | bash
boost init --accept-terms

Usage:

boost graph init # within repositories to index on demand