tshore2004/branchcontext

Branchcontext is a tool that gives AI coding assistants a memory that is tied to a specific Git branch rather than to a single chat session. Every time an AI coding session works on a branch, it can log what it did, why it made those choices, what risks it noticed, and what work is still left to do. The next time any compatible AI assistant, such as Claude Code or Cursor, starts working on that same branch, it can read back this history and pick up with proper context instead of starting from scratch. A key design choice is that there is no separate database or outside service involved. The history, called a ledger, lives directly inside the code repository itself, in a folder named .branchcontext. This means the memory travels with the code whenever it is committed, cloned, or shared, just like any other file in the project. The README is upfront that this is an early, version 0.1 scaffold. Right now only two actions work end to end: reading back a handoff, meaning the saved history for a branch, and logging a new handoff entry. These work over a communication method called MCP, which lets AI coding tools plug into external tools. Other planned features are described in separate planning documents included in the project but are not yet built. Setup involves installing the project's dependencies and building it with standard Node.js commands, then running a setup command that registers the tool with Claude Code by writing an entry into a configuration file. Each logged entry in the ledger is stored as a single line of JSON and includes details like the branch name, a base commit reference, a timestamp, which AI agent and model were used, a short summary, any decisions made, risks flagged, remaining to-dos, and which files were touched.