Using SnapBack with Claude Code

SnapBack integrates directly with Claude Code through the Model Context Protocol (MCP). This isn’t just detection—Claude can query SnapBack’s local risk engine before making changes.

Two Levels of Integration

Level 1: Automatic Detection (Free)

Like Cursor and Copilot, SnapBack automatically detects Claude Code activity:

• Recognizes Claude Code edit patterns
• Creates snapshots before changes
• Enables one-click restore

Level 2: MCP Integration (Pro)

With MCP enabled, Claude becomes aware of SnapBack:

CopyYou: "Refactor the auth module"

Claude: [Checks with SnapBack]
SnapBack: "High-risk operation. auth.ts has caused 3
          restores in the past week. Snapshot created."

Claude: "I've noted this is a sensitive area. Proceeding
         carefully with a snapshot ready..."

Setting Up MCP Integration

Step 1: Install the SnapBack CLI

Copynpm install -g @snapback/cli

Step 2: Configure MCP

Copysnapback mcp install

This adds SnapBack to your Claude Code MCP configuration.

Step 3: Restart Claude Code

Claude Code will now query SnapBack before risky operations.

What MCP Enables

MCP Tools Available

When MCP is enabled, Claude Code can use these tools:

snapback_analyze_risk

Claude asks: “Is this operation risky?”

Copy{
  "files": ["src/auth.ts", "src/config.ts"],
  "operation": "refactor"
}

SnapBack responds with risk score and historical context.

snapback_create_snapshot

Claude requests a snapshot before proceeding:

Copy{
  "message": "Before auth refactor",
  "files": ["src/auth.ts"]
}

snapback_check_history

Claude asks about past issues:

Copy{
  "file": "src/auth.ts"
}

SnapBack responds with restore history for that file.

Privacy Note

Next Steps

• Set up local MCP
• View MCP tools reference
• Explore MCP capabilities