How jeyaseelan6/tripmate-ai---a-multi-agent-travel-planner works
TripMate AI is a web-based, multi-agent travel planner. A user submits a natural-language request in the browser; the FastAPI endpoint invokes the LangGraph workflow, which validates and selects travel specialists, gathers research, builds an itinerary, and returns a polished answer. The browser displays the answer and supports copying or PDF export. The workflow also documents budget and human-approval state, though their execution paths are not visible in the excerpts. Relationships below use shown calls or README workflow descriptions; unsampled routing is omitted.
Main entry points: backend.py Guardrail and supervisor · mcp_client.py MCP tool client · app.py FastAPI endpoint · script.js Travel interface
Components 19
Web experience
- Travel interface
static/script.js - Planner page
templates - FastAPI endpoint
app.py
Travel workflow
- Workflow runner
backend.py - Guardrail and supervisor
backend.py - Flight agent
backend.py - Hotel agent
backend.py - Weather agent
backend.py - Budget and approval
backend.py - Itinerary agent
backend.py - Final response agent
backend.py
Travel research
- MCP tool client
mcp_client.py - Flight lookup
tools/flight_tool.py - Hotel web search
tools/tavily_tool.py - Weather MCP server
custom_weather_mcp_server.py
Conversation state
- PostgreSQL checkpoint store
Other components
- Traveler
- Travel tool services
- Groq LLM
Connections 20
- Traveler → Travel interface: submits request
static/script.js - Travel interface → FastAPI endpoint: POSTs request
static/script.js - FastAPI endpoint → Planner page: renders page
app.py - FastAPI endpoint → Workflow runner: invokes runner
app.py - Workflow runner → Guardrail and supervisor: starts workflow
backend.py - Guardrail and supervisor → Flight agent: selects agents
backend.py - Guardrail and supervisor → Hotel agent: selects agents
backend.py - Guardrail and supervisor → Weather agent: selects agents
backend.py - Guardrail and supervisor → Budget and approval: may select
backend.py - Guardrail and supervisor → Itinerary agent: selects agents
backend.py - Flight agent → MCP tool client: calls tools
backend.py - Hotel agent → MCP tool client: searches hotels
backend.py - Weather agent → MCP tool client: requests weather
backend.py - MCP tool client → Travel tool services: queries services
mcp_client.py - MCP tool client → Weather MCP server: starts server
mcp_client.py - Weather MCP server → Travel tool services: requests forecasts
custom_weather_mcp_server.py - Itinerary agent → Final response agent: provides itinerary
backend.py - Final response agent → Groq LLM: generates response
- Workflow runner → PostgreSQL checkpoint store: checkpoints state
backend.py - FastAPI endpoint → Travel interface: returns results
app.py