context-sync
Local persistent memory store for LLM applications including continue.dev, cursor, claude desktop, github copilot, codex, antigravity, etc.
claude mcp add --transport stdio intina47-context-sync npx -y @context-sync/server
How to use
Context Sync provides a local-first memory layer to accelerate AI development tools by exposing them through MCP (Model Context Protocol). After installing the server globally, it auto-configures within supported tools so they can discover and use a shared, persistent context and memory store. You can access a variety of integrations across popular editors and tools, including Claude Desktop, Cursor, VS Code with Copilot, Continue.dev, Codex CLI, Claude Code, and more. Use the documented tool references to recall, read, and remember context during your AI workflows, and rely on the local database and configuration to persist data across sessions. The README also points to a suite of docs (TOOLS, DATA, CONFIG, TROUBLESHOOTING) for deeper guidance on capabilities and troubleshooting.
How to install
Prerequisites:
- Node.js and npm installed on your system
- Administrative (sudo) access for global installation if needed
Steps:
- Install the Context Sync MCP server globally:
npm install -g @context-sync/server
- After installation, the server auto-configures with your connected tools. If auto-config fails, consult docs/ TROUBLESHOOTING.md and docs/CONFIG.md. If you installed locally (non-global), auto-config will not run automatically.
Optional Notion setup (not required for core operation):
context-sync-setup
# or
npx context-sync-setup
Verify installation and explore tools via the supported platforms list:
- Claude Desktop
- Cursor
- VS Code + Copilot
- Continue.dev
- Zed
- Windsurf
- Codeium
- TabNine
- Codex CLI
- Claude Code
- Antigravity (Google Gemini IDE)
Additional references for usage and configuration are available in docs (TOOLS.md, DATA.md, CONFIG.md, TROUBLESHOOTING.md).
Additional notes
Environment and configuration tips:
- Data is stored locally by default, e.g., the database at ~/.context-sync/data.db and configuration at ~/.context-sync/config.json. You can override the database path with CONTEXT_SYNC_DB_PATH or the --db-path option when starting the server.
- The install status is tracked at ~/.context-sync/install-status.json.
- When used inside a git repository with set_project, Context Sync can install git hooks (post-commit, pre-push, post-merge, post-checkout) labeled as Context Sync Auto-Hook with backups of existing hooks.
- If you want to customize the runtime, see docs/DATA.md for storage options and docs/CONFIG.md for manual configuration.
- The Tools reference is available in docs/TOOLS.md for understanding the capabilities and commands exposed by each integration.
Related MCP Servers
octocode
Semantic code searcher and codebase utility
ummon
The semantic layer for software engineering: Connect code to meaning, build on understanding
dbt-core
dbt Core MCP Server: Interact with dbt projects via Model Context Protocol
memcp
A persistent memory MCP server for Claude Code - Recursive Language Model integration for Claude Code inspired by MIT CSAIL paper
qdrant -pi5
Persistent semantic memory for AI agents on Raspberry Pi 5 — local Qdrant + MCP, no cloud, ~3s per query
classmcp
MCP server for AI-assisted CSS development. 77% token savings with semantic class patterns. Supports Tailwind, Bootstrap, UnoCSS, Tachyons.