sem

Why we included this project

Most version control tools still present changes as hunks of text, which makes it hard to tell what a commit actually did. sem parses your code with tree-sitter across 28 languages and reports changes at the level of real entities instead: the function that was added, the method that disappeared, the class that picked up a new dependency. It stays a thin layer on top of git, so you can adopt it in any existing repository without configuration. That entity-level view is what makes its blame and cross-file impact analysis useful; you can see which upstream edits touch code you own and which tests are affected, rather than hunting through shifted line numbers. There is also an MCP server for coding agents, so an agent can reason about code in structural terms instead of raw text.