trishul-mcp

Deterministic, repository-scale semantic code intelligence Model Context Protocol (MCP) server for AI coding agents (Claude Code, Cursor, Windsurf).

Overview

trishul-mcp connects AI coding assistants directly to a persistent, topological code intelligence graph. Instead of burning your model's context window re-reading thousands of raw files or guessing caller hierarchies with naive regex, Trishul provides deterministic answers with verified certainty:

• Callers (get_callers): Who directly or indirectly invokes a symbol?
• Blast Radius (get_blast / get_impact): What breaks upstream if this function or interface changes?
• Execution Paths (get_path): What is the optimal, deterministic call path between Symbol A and Symbol B?
• Dependency Hubs (get_dependency_hubs): What are the critical architectural bottlenecks and chokepoints?

Installation

pip install trishul-mcp

(Or install via pip install llmgraph-in)

MCP Client Configuration

Add to your claude_desktop_config.json, Cursor MCP settings, or Windsurf configuration:

{
  "mcpServers": {
    "trishul": {
      "command": "trishul-mcp",
      "env": {
        "TRISHUL_API_KEY": "llmg_live_...",
        "TRISHUL_API_URL": "https://api.llmgraph.in/v1"
      }
    }
  }
}

Note: LLMGRAPH_API_KEY and LLMGRAPH_API_URL are also supported for backward compatibility.

License

Apache-2.0. Developed by Brainfoods Studio.