BranchMind

AI-powered Git branch advisor for VS Code — local-first, privacy by architecture.

BranchMind watches your repository and suggests branch names, flags stale branches, detects naming inconsistencies, and warns you before switching away from dirty working trees — all without sending a single byte to the cloud.

Features

Branch health score

A 0–100 health score is computed from your repository's branch state — stale branches, naming convention drift, and untracked remotes all affect the score. The score is cached and refreshed only when new commits appear, so it's always fast.

AI-powered suggestions (local inference)

When a local LLM is running (Ollama, LM Studio, Jan.ai, or llama.cpp), BranchMind sends your recent diff (±15 lines of context, capped at 8 000 chars) plus staged file list to the model and shows commit message drafts and branch rename suggestions directly in the sidebar.

Privacy guarantee: All inference happens on your machine. No API keys, no telemetry, no cloud calls.

Automatic secret redaction

Before any diff reaches the model, 14 regex patterns strip credentials: AWS AKIA keys, OpenAI sk-, GitHub PATs, Slack tokens, Firebase private keys, Supabase JWTs, Stripe and Razorpay keys, generic SECRET=value assignments, connection strings, and non-localhost IPs.

Rule-based fallback

No local model running? BranchMind's deterministic rules engine still gives you branch naming suggestions and convention checks at zero latency.

Stale branch manager

For each stale branch (no commits in 30+ days) the panel shows:

• Merge — fast-forward merge into current branch
• Delete — confirmation-gated local branch deletion
• Revive — opens a terminal with git checkout -b <name>-revived

Branch-switch guard

fs.watch('.git/HEAD') detects branch switches the instant they happen. If you have uncommitted changes, BranchMind offers to stash them with a single click.

Monorepo support

BranchMind detects Turborepo, Nx, Lerna, and pnpm workspace layouts and scopes branch analysis to the active editor's package root.

Requirements

• VS Code 1.85 or later
• A Git repository (.git folder must exist in the workspace root)
• Optional: Ollama, LM Studio, Jan.ai, or any llama.cpp-compatible local server for AI suggestions

Getting started

1. Install the extension.
2. Open any Git repository in VS Code.
3. Click the BranchMind icon in the activity bar.
4. (Optional) Start a local LLM server and run BranchMind: Select Local Model from the Command Palette (Ctrl+Shift+P) to enable AI suggestions.

Commands

Extension settings

BranchMind stores per-project config in .branchmind/config.json inside your repository. This file is not added to .gitignore by default so project keywords are shared with your team. If you ever add an apiKey field, BranchMind automatically appends the config path to .gitignore.

Privacy

• No outbound HTTP calls except to localhost (local LLM endpoints).
• No telemetry or usage tracking of any kind.
• Secret redaction runs before every inference call — credentials in diffs are replaced with [REDACTED].

License

MIT — see LICENSE.