Skip to content
IntelliRag

MCP Tools

IntelliRag provides 20 MCP (Model Context Protocol) tools that give AI coding assistants deep, structured knowledge about your codebase. MCP is an open protocol that lets AI assistants call external tools to retrieve context, and IntelliRag uses it to surface code intelligence - symbols, call graphs, dependencies, technical debt, and more - directly in your development workflow.

How it works

Section titled “How it works”

The IntelliRag MCP server runs alongside your AI assistant as a local process. When you ask a question about your codebase, the assistant automatically selects the right tool, calls the IntelliRag API, and receives structured context that grounds its response in real indexed data rather than guesses.

┌────────────────┐     ┌───────────────┐     ┌──────────────┐
│  AI Assistant  │────>│  MCP Server   │────>│ IntelliRag   │
│  (Claude, etc) │<────│  (irag-mcp)   │<────│ API Server   │
└────────────────┘     └───────────────┘     └──────────────┘
     stdio transport        HTTP/API key

No manual invocation is needed. Your AI assistant decides when to call tools based on the context of your question.

Tool categories

Section titled “Tool categories”

IntelliRag’s 20 tools are organized into five categories.

Understanding

Section titled “Understanding”

Semantic search, symbol lookup, module summaries, and pattern detection. These tools help the assistant understand what your code does.

Understanding Toolssemantic_search, symbol_lookup, module_summary, pattern_search
Section titled “Navigation”

Call graphs, dependency graphs, data flow tracing, and entry point mapping. These tools help the assistant trace how code connects.

Navigation Toolscall_graph_query, dependency_graph, data_flow_trace, entry_point_map

Impact Analysis

Section titled “Impact Analysis”

Change impact analysis, dead code detection, complexity hotspots, git archaeology, and technical debt indexing. These tools help the assistant assess risk before changes.

Impact Analysis Toolschange_impact, dead_code_detector, complexity_hotspots, git_archaeology, tech_debt_index

Integration

Section titled “Integration”

API contracts, database schemas, event catalogs, configuration resolution, and auth policy mapping. These tools help the assistant understand how systems integrate.

Integration Toolsapi_contract_fetch, schema_browser, event_catalog, config_resolver, auth_policy_map

Generation

Section titled “Generation”

Test scaffolding and upgrade advisory. These tools help the assistant generate useful output grounded in your codebase structure.

Generation Toolstest_scaffold_gen, upgrade_advisor

How tools are called

Section titled “How tools are called”

You do not need to invoke MCP tools manually. When you ask your AI assistant a question, it examines the available tools and selects the ones most relevant to your request. For example:

  • “What does the UserService class do?” triggers symbol_lookup and semantic_search
  • “What would break if I change this method?” triggers change_impact and call_graph_query
  • “Show me all HTTP endpoints” triggers entry_point_map
  • “Find technical debt in the auth module” triggers tech_debt_index and pattern_search

The assistant may call multiple tools in sequence to build a complete answer. All tool responses are formatted as structured text optimized for LLM consumption - not raw JSON.

Next steps

Section titled “Next steps”
MCP Server SetupConfigure the MCP server for your AI assistant
QuickstartGet IntelliRag running with your codebase in 5 minutes