HomeMCP Tools

AI Integration

21 AI Tools. Zero Hallucination.

The Model Context Protocol (MCP) is USB for AI — a universal plug that lets any AI assistant talk to any data source. Nomik exposes 21 graph-powered tools that give your AI precise codebase intelligence.

How It Works

🤖

AI Client

Cursor / Windsurf / Claude

MCP stdio

↓ MCP Protocol ↓

⚡

Nomik MCP

21 tools exposed

Cypher

↓ Graph Query ↓

🧬

Neo4j Graph

752+ nodes, 1850+ edges

The Difference

Without Nomik

✗AI searches for "checkout" → 53 files returned
✗Context window overflows, deep relationships truncated
✗AI hallucinates — fills gaps with plausible-but-wrong info
✗Every conversation starts from zero (no memory)
✗AI misses the cron job that also depends on your function

With Nomik MCP

✓AI calls nm_impact → 5 precise dependents returned
✓Graph traversal follows exact call chains and dependencies
✓Real data from Neo4j — no guessing, no hallucination
✓Persistent graph survives across sessions and restarts
✓Graph reveals: CronJob:dailySync → CALLS → syncToStripe()

All 21 Tools

Every tool is project-scoped via NOMIK_PROJECT_ID and accepts an explicit project parameter for cross-project queries.

nm_search

Search for nodes by name, path, or pattern. Supports wildcards and case-insensitive matching.

Example prompt

"Find all auth-related functions"

nm_impact

Downstream impact analysis with confidence filtering. Which nodes break if a symbol is modified.

Example prompt

"What breaks if I change parseFile?"

nm_trace

Shortest dependency chain between two symbols (names only).

Example prompt

"Trace the chain from scanCommand to neo4j"

nm_path

Detailed path between two entities with node types and edge types at each step.

Example prompt

"How does parseFile connect to neo4j?"

nm_context

Rich context for a file or function: contains, calls, calledBy, imports, inheritance.

Example prompt

"Give me context for graph.service.ts"

nm_health

Full health metrics: dead code, god objects, god files, duplicates, all 17 node type counts.

Example prompt

"Are there any dead code or god objects?"

nm_explain

Full symbol explanation: type, file, callers, callees, edge counts, container context.

Example prompt

"Explain the createGraphService function"

nm_db_impact

DB table/column impact: who reads from or writes to a specific table or column.

Example prompt

"Who writes to users.email?"

nm_changes

Recently modified nodes by updatedAt timestamp.

Example prompt

"What changed in the last hour?"

nm_onboard

Full codebase briefing: stats, languages, DB tables, APIs, env vars, high-risk functions.

Example prompt

"Give me a codebase overview"

nm_communities

Functional cluster detection. Groups of code that frequently call each other.

Example prompt

"Show me the code communities"

nm_flows

Execution flow tracing from entry points (routes, events, queues) through the call graph.

Example prompt

"Trace the request lifecycle for /api/checkout"

nm_wiki

Structured documentation data: file index, top functions, health report, service links.

Example prompt

"Generate documentation for this module"

nm_guard

Quality gate check: dead code, god files, duplicates against configurable thresholds.

Example prompt

"Does the codebase pass quality gates?"

nm_rules

Evaluate 9 architecture rules + custom Cypher rules from .nomik/rules.yaml.

Example prompt

"Run architecture rules check"

nm_rename

Graph-aware rename impact: all callers, importers, and affected files for a symbol.

Example prompt

"What files are affected if I rename parseFile?"

nm_diff

Architecture drift between two git SHAs: new/removed files, functions, and call edges.

Example prompt

"What changed architecturally since last release?"

nm_test_impact

Find test files affected by changing a symbol or file. Graph-based, no grep needed.

Example prompt

"Which tests should I run after changing parseFile?"

nm_audit

Dependency vulnerability check with graph blast radius — which files import vulnerable packages.

Example prompt

"Are there any vulnerable dependencies?"

nm_service_links

Cross-service dependencies: producer/consumer pairs for queues, events, and API calls.

Example prompt

"Show cross-service connections"

nm_projects

List all projects tracked in the knowledge graph with metadata.

Example prompt

"What projects does NOMIK track?"

Live Example: Impact Analysis

When you ask Cursor: "What happens if I modify updateUserProfile?"

AI sends tool call

{
  "tool": "nm_impact",
  "arguments": {
    "symbolId": "updateUserProfile",
    "depth": 3
  }
}

Nomik returns graph data

[
  { "name": "PUT /api/users/:id",
    "type": "Route", "depth": 1 },
  { "name": "syncToStripe",
    "type": "Function", "depth": 1 },
  { "name": "users [DBTable]",
    "type": "DBTable", "depth": 1 },
  { "name": "sendWelcomeEmail",
    "type": "Function", "depth": 2 }
]

IDE Setup

Cursor

setup

$ nomik setup-cursor

✔ .cursor/mcp.json created

Auto-creates config with Neo4j credentials and project ID.

Windsurf

setup

$ nomik setup-windsurf

✔ mcp_config.json created

Targets ~/.codeium/windsurf/mcp_config.json by default.

Claude Desktop

setup

$ nomik setup-claude

✔ claude_desktop_config.json updated

Auto-configures Claude Desktop's MCP settings.

ℹ

stdio mode = zero config. Cursor and Windsurf launch the MCP server on demand. You don't need nomik serve for AI integration — only for the visualization dashboard.

Transport Options

Transport	Use Case	How
stdio	Cursor, Windsurf, Claude Desktop (local)	Process spawned by the client — recommended
SSE	Remote access, web dashboards	HTTP server on configurable port
Streamable HTTP	Production / multi-client	Stateless HTTP with streaming

← Home

→

Vision