How harish-ai-engineer/langgraph-agent- works
This is a customer-support assistant with a LangGraph conversational workflow and a separate MCP interface to the same tools. A user submits a question through the CLI or an MCP client; the graph consults a configured language model, dispatches tool calls when needed, and returns a response. The tools provide arithmetic, sample weather, and support-article lookup. Source excerpts show the graph and tool dispatches; README documentation establishes the MCP stdio entry point. The weather and knowledge data shown are local sample data, not external services.
Main entry points: graph.py Conversation graph · server.py MCP server · main.py CLI entry point · tools.py Calculator
Components 13
Agent runtime
- CLI entry point
main.py - Conversation graph
agent/graph.py - System prompt
agent/prompts.py - Model config
agent/config.py - Tool dispatch
agent/graph.py
Support tools
- Calculator
agent/tools.py - Weather lookup
agent/tools.py - Knowledge lookup
agent/tools.py
MCP interface
- MCP server
mcp_server/server.py - stdio transport
mcp_server/server.py
Other components
- User
- Language model
- MCP client
Connections 16
- User → CLI entry point: asks
main.py - CLI entry point → Conversation graph: invokes
main.py - Conversation graph → CLI entry point: returns response
main.py - Conversation graph → System prompt: formats with
agent/graph.py - Conversation graph → Model config: reads settings
agent/graph.py - Conversation graph → Language model: invokes model
agent/graph.py - Language model → Conversation graph: returns message
agent/graph.py - Conversation graph → Tool dispatch: routes tool calls
agent/graph.py - Tool dispatch → Calculator: dispatches
agent/graph.py - Tool dispatch → Weather lookup: dispatches
agent/graph.py - Tool dispatch → Knowledge lookup: dispatches
agent/graph.py - MCP client → stdio transport: connects over stdio
README.md - stdio transport → MCP server: serves requests
mcp_server/server.py - MCP server → Calculator: calls on request
mcp_server/server.py - MCP server → Weather lookup: calls on request
mcp_server/server.py - MCP server → Knowledge lookup: calls on request
mcp_server/server.py