Context Sync

The core of Context Sync is a remember and recall pair backed by a local SQLite database. You store decisions, constraints, and caveats with a type tag, and any connected assistant can pull them back later. That means a choice you logged last month still shows up when it's relevant, without you retyping it. No copy-paste ritual. No guessing what you decided in March. The database lives at ~/.context-sync/data.db by default, and you can move it with the --db-path flag or the CONTEXT_SYNC_DB_PATH environment variable.

set_project establishes which codebase you're working in, then recall rebuilds the context window from stored data. The recommended flow is to set the project at the start of each session and recall right after. It's a small habit that saves a lot of typing, and it works the same across every assistant that speaks MCP. Run it once. Forget about it.

Instead of dumping directory trees into a chat, Context Sync exposes structure, search, and read_file as discrete tools. You ask for a folder depth or a filename pattern and get a focused result. This keeps context windows lean, which matters when you're feeding a model a large project and want the relevant parts, not everything. Why does that matter? A bloated context window buries the signal you actually need, and every wasted token is one the model can't spend on your actual question.

When you run set_project inside a git repository, the server can install hooks for post-commit, pre-push, post-merge, and post-checkout. Each hook is marked "Context Sync Auto-Hook," and any existing hooks get backed up first. The git tool then reports status, hotspots, coupling, blame, and analysis, so you can see which files churn most often and which ones tend to change together, even when nobody wrote down that relationship anywhere.

If your team lives in Notion, the setup wizard (context-sync-setup) connects a search and read interface. Support is intentionally read-only: you can search pages and read a page by ID, but the tool can't write back. That's a sensible default for a memory server that shouldn't be silently editing your docs. Anyone who's watched a bot mangle a wiki page will appreciate the restraint.

Context Sync connects to Claude Desktop, Cursor, VS Code with GitHub Copilot, Continue.dev, Zed, Windsurf, Codeium, TabNine, Codex CLI, Claude Code, and Antigravity. Auto-config runs on global install and wires up most of them, and the rest follow the manual guide in the project docs, which walks through editing each tool's MCP config file by hand. Cross-platform support covers macOS, Linux, and Windows on x64 and arm64. Broad coverage. Very little lock-in.