arcanon

Ligamen

Your AI coding agent doesn't know what it's about to break.

Claude Code is powerful — it writes, refactors, and ships code fast. But it operates blind to what lives in your other repositories. It doesn't know that the endpoint it just renamed is called by three downstream services, or that the schema it changed is shared across your entire platform.

Ligamen gives Claude Code a service dependency graph that spans all your repositories. Before changes are made, Claude can see which services are connected, trace the blast radius of a change, and catch drift across API contracts, shared types, and dependency versions — so your AI agent stops introducing cross-service bugs.

What Ligamen Does

Maps your architecture. Ligamen scans your linked repositories with Claude agents to build an interactive service dependency graph — services, libraries, infrastructure, external actors, and every connection between them. Explore it visually in an interactive graph UI at http://localhost:37888.

Shows what breaks before it breaks. Run /ligamen:cross-impact and Ligamen traces dependencies through your service graph, flagging every downstream service affected by your changes — ranked by severity. Your AI agent can check this automatically via MCP tools before making any modification.

Catches drift across repos. Dependency versions out of sync? Type definitions that diverged? OpenAPI specs that don't agree? /ligamen:drift finds the inconsistencies before they become production incidents.

Keeps your code clean automatically. Every file Claude edits gets auto-formatted and auto-linted. Sensitive files like .env, lock files, and credentials are protected from accidental writes. No configuration needed.

Quick Start

claude plugin marketplace add https://github.com/chilleregeravi/ligamen
claude plugin install ligamen@ligamen --scope user

That's it. Ligamen works with zero configuration — hooks activate immediately, and commands are available in every Claude Code session.

Build your first service map:

/ligamen:map

See what your changes affect:

/ligamen:cross-impact

Check for drift across repos:

/ligamen:drift

Commands

Command	What it does
/ligamen:map	Scan repos and build service dependency graph
/ligamen:map view	Open the graph UI without re-scanning
/ligamen:cross-impact	Trace blast radius of current changes across services
/ligamen:drift	Find version mismatches, type drift, and API spec divergence
/ligamen:quality-gate	Run lint, format, test, and typecheck for your project

See Commands for full usage and options.

Automatic Behaviors

Ligamen runs these in the background on every Claude Code session with zero setup:

• Auto-format — formats every file Claude edits (Python, Rust, TypeScript, Go, JSON, YAML)
• Auto-lint — runs your project's linter and surfaces issues to Claude
• File guard — blocks writes to .env, lock files, credentials, and generated directories
• Session context — detects your project type and auto-starts the graph worker if configured

See Automatic Behaviors for details and how to disable individual behaviors.

Graph UI

After scanning, open http://localhost:37888 to explore your service architecture visually — layered layout, boundary grouping, protocol-differentiated edges, subgraph isolation, blast radius highlighting, what-changed overlay, filtering by protocol/language/boundary, and PNG export.

See Service Map for the full feature set.

MCP Server

After building your first map, add the Ligamen MCP server so every Claude agent — not just the session that ran the scan — can check impact before making changes:

{
  "mcpServers": {
    "ligamen-impact": {
      "type": "stdio",
      "command": "node",
      "args": ["<path-to-ligamen>/plugins/ligamen/worker/mcp/server.js"]
    }
  }
}

Add this to your Claude Code MCP settings (typically ~/.claude/settings.json under "mcpServers").

This exposes 8 tools to all Claude sessions: impact_query, impact_changed, impact_graph, impact_search, impact_scan, drift_versions, drift_types, and drift_openapi.

Configuration

Ligamen works with zero configuration. For customization, see Configuration — linked repos, service boundaries, ChromaDB semantic search, environment variables, and machine settings.

Documentation