Claude Context Sync

Created By
ybin454814 days ago
Share context, detect file conflicts, and send messages across parallel Claude Code sessions
Overview

claude-context-sync

npm License: MIT

MCP server that automatically summarizes and shares context across parallel Claude Code sessions.

When you run multiple Claude Code sessions on different tasks, each session is unaware of what the others are doing. claude-context-sync watches your session logs, generates structured summaries on demand, and provides real-time file conflict detection and cross-session messaging.

Features

  • Session context sharing — any session can see what others are working on
  • File conflict detection — real-time alerts when multiple sessions edit the same file
  • Symbol-level conflicts — detects shared function/class/interface edits (TypeScript, Python, Swift, Go, Rust)
  • Cross-session messaging — send messages between sessions without switching terminals
  • Lazy evaluation — zero cost when tools aren't called
  • Auto-cleanup — tracked data is removed when sessions end

Installation

npm install -g claude-context-sync
context-sync init

That's it. The next time you open Claude Code, the MCP server starts automatically. The init command also registers a conflict detection hook.

MCP Tools

list_sessions

Returns active Claude Code sessions with their current status, task summary, and any file conflicts.

get_session_context

Returns the full structured summary for a specific session. Automatically re-summarizes if stale. Includes conflict information.

Parameters: sessionId (string)

get_all_changes

Returns changed files and decisions across all sessions, with optional project filtering.

Parameters: project (string, optional)

send_message

Sends a message to another session. Omit targetSessionId to broadcast to all sessions in the same project.

Parameters: fromSessionId (string), message (string), project (string), targetSessionId (string, optional)

get_messages

Returns message history for a session.

Parameters: sessionId (string), unreadOnly (boolean, optional)

resolve_conflicts

Clears resolved files from the conflict tracking list.

Parameters: sessionId (string), files (string[])

How It Works

Context Sharing

[Always running] Watcher: detects .jsonl changes → sets stale flag (zero cost)
[On tool call]   MCP Tool invoked → summarizes only stale sessions
                 Extractor → Generator (claude -p) → Store → Response

Conflict Detection

[Always running] Watcher: detects .jsonl changes
                 FileTracker: parses Write/Edit tool_use events
                 SymbolResolver: identifies enclosing function/class/type
                 Writes per-session touched files

[Every message]  Hook (UserPromptSubmit): checks for new conflicts
                 Alerts Claude once per new conflict → Claude notifies user

Messaging

Session A: send_message → writes to ~/.claude/context-sync/messages/{targetId}/
Session B: Hook detects unread message → injects into AI context
           Claude reads and relays to user → marks as read

Architecture

Single MCP server process — no separate daemon. Claude Code launches it automatically via stdio transport.

src/
├── mcp/server.ts              # MCP server + 6 tools
├── watcher/
│   ├── watcher.ts             # File watching (chokidar)
│   ├── scheduler.ts           # Hybrid summarization strategy
│   ├── file-tracker.ts        # Real-time file edit tracking
│   └── symbol-resolver.ts     # Pattern-based symbol detection
├── summarizer/
│   ├── extractor.ts           # .jsonl parsing + stratified sampling
│   └── generator.ts           # claude -p invocation
├── messaging/
│   └── messenger.ts           # Cross-session message passing
├── store/store.ts             # Local summary cache
├── cli.ts                     # CLI (init + serve + hook registration)
└── types.ts                   # Shared types

Summarization Strategy

  • Incremental: updates existing summary with new messages (default)
  • Full: regenerates from scratch (after 4 incremental updates)
  • Stratified sampling: for large sessions (500+ messages), samples head + middle + tail
  • Cached: returns stored summary if nothing changed (zero cost)

Supported Languages (Symbol Detection)

LanguageDetected Symbols
TypeScript/JavaScriptfunction, class, interface, type, enum, const
Pythondef, class
Swiftfunc, class, struct, protocol, enum, extension
Gofunc, type (struct/interface)
Rustfn, struct, enum, trait, impl

Unsupported file types fall back to file-level conflict detection.

Requirements

  • Node.js 20+
  • Claude Code CLI (claude command available)
  • jq (for conflict detection hook)

Feedback

We'd love to hear from you! If you have feature requests, bug reports, or just want to share how you use claude-context-sync:

Development

git clone https://github.com/ybin4548/claude-context-sync.git
cd claude-context-sync
npm install
npm run build
npm test

License

MIT

Server Config

{
  "mcpServers": {
    "claude-context-sync": {
      "command": "npx",
      "args": [
        "-y",
        "claude-context-sync",
        "serve"
      ]
    }
  }
}
Project Info
Created At
14 days ago
Updated At
14 days ago
Author Name
ybin4548
Star
-
Language
-
License
-
Category

Recommend Servers

View All
Bring your real authenticated browser session to AI coding agents. Local-first MCP server + Chrome MV3 extension. No cloud. No telemetry.
@Cubenest

peek records the user's actual logged-in browser (DOM via rrweb, console events, network metadata, optional response bodies via opt-in Deep capture) through a Chrome MV3 extension. The extension ships events through a native-messaging stdio bridge to a local MCP server (peek-mcp), which persists them to a SQLite database at ~/.peek/sessions.db. AI coding agents (Claude Code, Cursor, Cline, Windsurf) read sessions from the database via 10 MCP tools: Tool What it does list_recent_sessions List recently recorded sessions (id, origin, ts, event count). get_session_summary LLM-readable narrative summary of a session. get_session_console_errors Console errors recorded in a session. get_session_network_errors Failed/notable network requests in a session. get_user_action_before_error Last N user actions before a console error. generate_playwright_repro Generate a runnable Playwright test from a session. get_dom_snapshot Reconstruct the DOM at a given timestamp. query_dom_history Timeline of attribute/text changes for a selector. request_authorization Side-panel consent for write actions (Level 3). execute_action Dispatch a UI action (gated by permission level + destructive blocklist). Why local-first matters Every other "browser session for AI" tool ships to a vendor cloud. peek's SQLite + extension live on the user's machine — no remote endpoints, no telemetry. The privacy policy (docs/peek/PRIVACY_POLICY.md) is the source of truth. Install # 1. Add the MCP server to Claude Code claude mcp add peek -- npx -y @peekdev/mcp # 2. Install the Chrome extension from the Chrome Web Store # (link added once the CWS listing is approved)

a day ago
Crevio

2 days ago